IT484 Cyber Forensics_Information Technology

‫اﻟﺠﺎﻣﻌﺔ‬
‫ﺔ‬ ‫اﻟﺴﻌﻮد‬
‫ﺔ‬ ‫وﻧ‬ ‫اﻻﻟ‬
‫اﻟﺠﺎﻣﻌﺔ‬
‫ﺔ‬ ‫اﻟﺴﻌﻮد‬
‫ﺔ‬ ‫وﻧ‬ ‫اﻻﻟ‬
26/12/2021

College of Computing and Informatics
Bachelor of Science in Information Technology Program
IT484: Wireless Sensor Networks

Module 1
Introduction to Wireless Sensor Networks (WSN)

Contents
1. Introduction
2. Background
3. Components of a wireless sensor node
4. Classiﬁcation of sensor networks
5. Characteristics of wireless sensor networks
6. Challenges of wireless sensor networks

Contents
1. Limitations in wireless sensor networks
2. Design challenges
3. Hardware architecture
4. Operating systems and environments
5. Examples of sensor nodes
6. Effect of infrastructure on the performance evaluation of WSNs
7. Hardware platforms
8. Software platforms

Weekly Learning Outcomes
1. Describe the components of a wireless sensor node.
2. Discuss the classiﬁcation of sensor networks.
3. Discuss the characteristics and challenges of wireless sensor
networks (WSNs).
4. Understand the design constraints for WSNs
5. Explain the hardware architecture, operating systems and
environments for WSNs
6. Understand the effect of infrastructure on the performance
evaluation of WSNs

References
Chapters 1 & 2
Principles of Wireless Sensor Networks, M. Obaidat and
S. Misra, Cambridge University Press, 2014.
(ISBN: 978-0-521-19247-7)

Introduction : What is Wireless Sensor Network
 A wireless sensor network (WSN) : wireless network that contains distributed
independent sensor devices
 Monitor physical or environmental conditions.
 Set of connected tiny sensor nodes- exchange information and data through
communication with each other.
 Nodes gather information on the environment such as temperature, pressure, humidity
etc. and send it to a base station.
 Data is sent from base station to a wired network or activates an alarm for an action.
 Differs from wired network : Small sources of energy equipped Sensors are components
of the network  limited resources

Components and Application of WSN
 Numerous small and inexpensive nodes,
 Location of the nodes is either static or dynamic – decided by the type of application it is
used.
 Resource constrained devices – components of WSN
 Monitoring – Main application of WSN
 Three classes of monitoring :
 entity monitoring: civil structures or the human body
 area monitoring, monitoring the environmental area alarms
 area-entity monitoring : vehicles on the highway, and monitoring movement of
objects

 Key significance of WSNs : contributed by the array of interconnected
sensor nodes
 A large scale WSN : Expected to provide reliability and self configuration.
Why WSN?

• Sensor nodes communicate with each other by means of a multi-hop scheme.
• Data acquired by the sensors transmitted to the base stations or Sink.
• The data is further processed and sent to respective network for analysis.
• Sensors are interconnected in small groups : Clusters
• Each cluster has head (leader) node.
• Communication among the nodes in a cluster is through the head node.
• Head node is responsible to communicate the gathered data to the base station.
How does WSN operate?

 Initially used in military applications.
 Deployed in many civilian applications
 Environment monitoring
 Industrial process monitoring
 Health care applications
 Road and highway trafﬁc control
 Smart homes and cities
 Ofﬁce automation.
Evolution and Application of WSN

Sensor and Types of Sensor Nodes
Hardware devices : sense the data and produce
response in a measurable form.
Small sized microelectronic sensing devices with
limited resources
Types : based on their operation
 physical sensors, thermal sensors, chemical sensors,
biological sensors, electromagnetic, optical, and acoustic
sensors.
Examples of Commercial sensors :
 BTnode, BEAN, COTS and DOT, MICA and KMote.
Major challenge : Energy source – short lived
battery power

• Sensor Node : central component of a WSN
• tiny device that senses its immediate environment and map or store the
information.
• Consists of 4 parts : Microcontroller, Transceiver, External Memory, Power
source
Components of a wireless sensor node

Basic Components of Sensor Node
Microcontroller: Miniature sized computer on a chip
@ capable of doing powerful tasks : controlling the functions of other devices
connected to it.
@ Comprises of microprocessor, a RAM memory, and associated peripherals.
Transceiver: used to send and receive data, and commands through radio
frequency.
@Usually use the industrial, scientiﬁc and medical (ISM) frequency bands.

Basic Components of Sensor Node – Contd..
External memory: Flash memories - small size and reasonable storage
capacity .
Holds User and a program data.
Size of the external memory depends on the application.
Power source: Batteries
Sources of power consumption : transmitting data, node
programming, sensing and data collecting, data processing, and
data communication

 Categorized the sensor network designs and protocol implementation based
on changing requirement
 Key distinguishing features are :
 Data sink(s)
 Sensor mobility
 Sensor resources
 Traffic pattern:
Classification of sensor networks

Data sink(s): crucial features of sensor networks is the characteristic of data
sink(s).
 Static node inside the network
 Mobile access point – gathers data periodically.
 Efficiency is based on data storage methods
Sensor mobility : Organization of the sensor nodes
 Stationary or mobile sensor nodes. Localizations services with mobility feature
Sensor resources : Memory capacity and processing conditions affect the protocol
implementations
Traffic pattern: density of the data traffic
 Event driven applications – produce data traffic only on the occurrence of the event
 Environmental monitoring : Constant data traffic produced

• Wireless sensor network taxonomy can be based on the following
dimensions:
1. Spatial resolution: Metric units of measurement - centimeters, meters, or
millimeters.
2. Latency: Time elapsed between the sensing of the event and data received at the
data sink - negligible, moderate, or high.
3. Coverage: Observable physical space range of the sensor - partial, full, or
redundant.
4. Control: Classes here can be external, central, or distributed.
5. Temporal resolution: measured in seconds.
6. User types: single, competitive, cooperative, and collaborative classes.

• Wireless sensor network taxonomy can be based on the following
dimensions:
7. Lifetime: simple with fixed duration, or complex with multiple phase-specific fixed
durations.
8. Bandwidth: episodic-small, episodic-large, continuous-small, and continuous-large
categories. Units of bandwidth can be bytes/episode or byte/second.
9. Sense of occurrence: single discrete-target, multiple discrete-targets, and single
distributed phenomena, and multiple distributed phenomena
• Some classify WSNs based on the following two concepts: (a) network
organization or structure, and (b) node fairness and capabilities

 Inexpensive, smart devices with many on-board sensors networked through
wireless links.
 Primarily, sensors are electrical, electronic, or electromechanical devices,
even though other kinds of sensors exist.
 Sensors can be direct or paired.
 Direct sensor : thermometer values obtained from reading the values indicated by
the device
 A paired sensor : converts analog signal to digital signal using an analog-to-digital
(A/D) converter.
 Applied in medicine, industry, environment, robotics, and military.
 Micro-Electro-Mechanic-Systems (MEMS) technology built sensors -
advances in material
Characteristics of wireless sensor networks

Characteristics of Sensor device
1. It should be responsive to the considered property.
2. It should be insensible to any other property.
3. It is desirable that the output signal of the sensor is exactly proportional to the value
of the measured characteristic.
4. It should have a reasonable lifetime.
5. It should not consume much power.

• Comprised of interconnection of sensing devices to observe different conditions and
environments: motion, pressure, temperature, sound, vibration, pollution etc. at
different sites.
• Sensing devices are tiny and low-cost deployed in large quantities.
• Limited battery life : Power saving (sleep) mode nodes or in handing out the sensor data.
• Main functionalities  sensing, communication, and computing.
• Categorized : addressing mode of the nodes  separately addressable ( individual data)
or aggregated data from a group of nodes
Characteristics of wireless sensor networks

Challenges of wireless sensor networks
1. Scalability
2. Power limitation.
3. Self-organization.
4. End objective.
5. Querying capability.
6. Interoperability
7. Cost
8. Transmission time.
9. Data compression.
10. Interference and environment.
11. Security

Inside a Wireless Sensor Node: Structure and Operations

Limitations in wireless sensor networks
 Sensor nodes : positioned in an unfriendly remote environments
 Major parameters that design decisions of protocols:
 restricted computational power and energy
 limited storage space in the nodes
 Eg: Lightweight security protocols, energy aware routing protocols

• Two types of information : Traffic and Signalling
• Trafﬁc: user-to-user information. Eg: raw data, voice, or video.
• Signaling: Information for operating. Eg: maintenance, security, or trafﬁc
routing control, among others.
• Common Signalling types: Per-trunk signaling, Common channel Signalling
Per-trunk signaling (PTS): Signaling and
voice elements are sent on the same facility
Common channel signaling (CCS):
Signaling and voice elements are sent on two
different split paths.
allows the voice component to be assembled
resources be saved individually
Eg: No voice signaling when the number is
busy

• Criticial tasks monitoring : military, industrial plan monitoring, forest ﬁre
monitoring, smart homes, and health care applications.
• Challenge : security constraints that should be dealt with at the early
stages of design.
• designing new security procedures is limited by the resources of the sensor
nodes.

• Among the major limitations of WSNs are the following:
1. Arbitrary topology
2. Energy limitations
3. Storage limitations
4. Limited computational power.
5. Unfriendly environment

• There are distinctive features for WSNs that make them unique. The
realization of a protocol for them must take into consideration the
properties of ad hoc networks, in addition to the following.
1. Lifetime restrictions due to the limited energy provisions of the nodes in the network.
2. The communication is unreliable due to the nature of the wireless transmission
medium.
3. The need to be small and with or without a human intervention.
4. The need for self-conﬁguration and fault tolerance.

Design Challenges
 Flexibility and redundancy: The WSN should be designed so that if a node breaks down
or loses its power then the remainder of the WSN should continue its operation without
interruption.
 Scalable and adaptable structural design: AWSN must be ﬂexible enough to allow
expansion.
 Adding more nodes should not affect routing and clustering operations. Adapteable
to the new topology
 improvized performance.
 Unreliability of the wireless transmission medium: Unreliable wireless transmission
medium
 Caused of unreliability : atmospheric noise, interference, scattering, reﬂection,
diffraction, hence the signal attenuates and bit error rate increases.

Design Challenges
 Real time: Applications in real-world environments,
Actions and data must be sent in real-time : need for real time protocols
Ensure efﬁciency of addressing the real-time nature of such operations
 Security and privacy:
Deployed in remote and hostile environments: highly prone physical attacks.
Secure WSN: Every component integrated with Security features.
Key issues - protect the WSN links from tampering and eavesdropping.

 Every sensor node : a radio transceiver, a small microcontroller, a storage
space, and power source.
 Size of a sensor node may vary in size: it can be as small as a peanut and as
big as a soda can.
 Cost of a sensor node varies from tens of cents to hundreds of dollars,
depending on the required operations in the node.
 The size and cost constraints on sensor nodes result in corresponding
constraints on node’s resources including memory, I/O, speed, and power.
Hardware architecture

The major elements of a wireless sensor node are:
(1) transceiver,
(2) Microcontroller
(3) memory device
(4) power source
(5) sensing element(s)

 Transceiver: combined functions of transmitter and receiver
 Modes of operation of a transceiver : transmitter, receiver, and idle/sleep modes
 Wireless transmission medium
 Infrared, radio-frequency, and optical ﬁber
 Utilize the ISM band.
 Communication based on radio-frequency (RF)
 Microcontroller: Performs important tasks needed to have proper operation of entire WSN.
 Processing data and controlling the operations of other elements in the sensor node.
 Memory device: Memory include on-chip memory of the microcontroller and an external ﬂash
memory.
 Flash memories - cost-effectiveness characteristics; high capacity at low cost.
 Memory type is divided into two major classes: (a) data memory, and (b) program memory

 Power source: Batteries or Capacitors.
Chargeable and non-rechargeable batteries.
Classified based on electrochemical material used for electrode: nickel-cadmium
(NiCd), nickel-zinc (NiZn), nickel metal hydride (Nimh) and lithium ion.
Latest sensor’s battery get charged from the Sun, heat or movement
 Sensing element(s): generate a signal proportional to the event or condition being
monitored or measured.
Sensed signal is typically converted to digital form using A/D converters.
Characteristics of a sensor node: small size, low power consumption, able to work
in high volumetric densities, adaptive to environment, and independent and able to
work unattended.
Types of sensors include: (1) Active sensors, (2) Passive and omnidirectional
sensors, and (3) Passive and narrow-beam sensors.

• The ﬁrst operating system that was speciﬁcally designed for WSNs is the
TinyOS.
Operating systems and environments
 Designed as event-driven programming model; not based on multithreading model.
 Programs in TinyOS are arranged into event handlers and tasks through run-to-
completion semantics.
 After an outer event such as an arrival of a packet or arrival of sensed data, the TinyOS
calls the suitable event service routine to run the event, which usually can be a read or
write task.
 Service-handling subroutines can schedule tasks that are planned by the TinyOS kernel
some time later.
 Programming language used to write system and application programs under TinyOS is
called nesC, which is a sort of C programming language

• Other operating systems that permit programming in C include SOS,
Contiki, BTnut, Nano-RK, and LiteOS.
 SOS is an event-driven-based operating system that supports loadable modules. It also
focusses on supporting dynamic memory management.
 The Contiki kernel is also event-driven- supports multi-threading
 includes proto-threads : offer a thread-like programming abstraction
 require small memory overhead.
 The BTnut is founded on cooperative multi-threading base and ordinary C code.
 Nano-RK, it is a real-time resource kernel that allows ﬁne-grained control.
 LiteOS offers UNIX-like construct and supports C

• Middleware : software that links various system software modules, or
application programs.
• Efficient middleswares are a key part of the WSNs
• The main schemes to design middleware for WSNs are based on:
1. Distributed database
2. Mobile agents
3. Event-based

• University of California–Berkeley and the Intel Berkeley Research Laboratory- a self-
organizing WSN composed of over 800 small low-power sensor nodes.
• 1 × 1.5 inch wireless sensor node was demonstrated.
• The core was a 4 MHz low-power microcontroller (ATMEGA 163) with 16 kB of ﬂash
instruction memory, 512 bytes of SRAM, A/D convertors, and basic I/O interfaces.
• A 256 kB EEPROM was used as secondary storage, and sensors, actuators, and a radio
network used as the I/O subsystem.
• The network employed a low-power radio (RF Monolithic T1000) running at 10 kbps.
• Any node in the system had four sensors: light, temperature, battery level, and radio
signal strength sensors and it could trigger two LEDs, direct the signal power of the
radio, send and receive signals.
Examples of sensor nodes

Effect of infrastructure on the performance evaluation of WSNs
 Structure of WSN : three-layer system
 Infrastructure: physical sensors’ properties and abilities, number of sensors, and
their launching approach.
 Networking protocol: This is in-charge of dissemination of the sensed data by
creating and maintaining paths between the sensors and the observer(s).
 Application: This is responsible for explaining the observer interests into explicit
network-level actions. Optimizations across the three levels are feasible in order to
enhance the overall performance of the network.

• Performance evaluation metrics : Delay, Correctness, Goodput, Scalability,
Energy effectiveness, Fault tolerance
 Delay: The time needed to get the samples to the analyst.
 Causes of latency : WSN congestion, and the duty cycle of the sensors
 Real-time trafﬁc - real-time voice or video
 Effect of delay : affect the correctness of data and accuracy of operation.
 Correctness : The accuracy of measurement made at the sensor.
 Specific to a physical transducer and environment.
 Data from multiple sensors – merged to produce accurate approximation of the place
of incidence.
 Goodput. Measure of the ratio of the total number of packets obtained by the
observer to the total number of packets transmitted by all the sensors.

 Scalability : Refers to whether increasing the number of nodes will provide
proportional improvement in the overall performance of the system.
 Energy effectiveness : measured by using various methods such as the consumed
energy within the network.
 Sensor nodes are run using batteries: Energy efficient protocols to extend their
lives, which means extending the life of the WSN.
 Fault tolerance. Refers to the degree at which the network can perform properly
even when it loses some nodes for one reason or another, such as physical damage
or running out of power.
 Replacing sensors is very difﬁcult; therefore, the WSN should be fault tolerant so
that non-disastrous breakdowns are veiled from the application.

• A micro-electro-mechanical system (MEMS) is important technology for
making tiny, inexpensive, and low-power sensor nodes.
• MEMS-based sensor devices provide a periphery that can sense, treat,
and/or direct the contiguous environment.
• MEMS-based sensors: devices constructed using very small electrical and
mechanical elements on a single chip.
• Sensors : Vital component in wireless devices, computer peripherals, hard-
disk drives, and smart portable electronic devices such as cell phones.
• The key beneﬁts of MEMS are low cost, low power requirements, small size,
integration, and high performance
MEMS technology

 MEMS technology: Reduce volume and cost of sensor nodes.
 To attain low power expenditure, it is essential to include awareness and
energy optimization in the hardware structure for sensor networks.
 Platforms for sensor node hardware can be categorized into three key
classes
 System-on-chip (SoC) sensor nodes
 Augmented general-purpose personal computers
 Dedicated sensor nodes
Hardware platforms

• Software platform : operating system
• Set of applications such as ﬁle organization, memory allotment, task
scheduling, and networking.
• Language platform - offers a library of elements to programmers.
• Example software platforms for WSNs such as Mote, TinyOS, and nesC,
among others
Software platforms

Module 2
Wireless Sensor Network Applications:
Overview and Case Studies

Contents
1. Target detection and tracking using WSNs
2. Contour and edge detection using WSNs
3. Types of applications

1. Explain target detection and tracking.
2. Discuss contour and edge detection.
3. Discuss the applications of WSNs.

References
Chapter 3
(ISBN: 978-0-521-19247-7)

Introduction
• Increased applications of WSNs over the recent years -> civilian and
military sectors
• Increasing applications of wireless sensor networks (WSNs) have motivated
the research as well as development communities
• Most WSNs are built for delay-tolerant and low-bandwidth applications =>
most research efforts focus on this latter paradigm, which is often called
terrestrial sensor networks.
• Network lifetime optimizations is the key challenge in any WSN
application.
• A WSN structure includes a gateway that provides connectivity with the
wired network.

Introduction
• In almost all the applications, WSNs are used to monitor physical processes
or quantities.
 Can be deployed for measuring the temperature, air pressure, chemical reactions,
movement of objects, vitals of the body, etc.
 This can help recover some important information of the system boundaries that
are often referred to as edges.
 The recognition of a boundary is essential in order to keep track of the edge of a
physical development.
 The problem of locating the boundary is the ﬁrst step for resolving the edge
detection.
 In digital processing, various methods exist for identifying the edge but they are not
easy to implement in WSNs as the sensor nodes are not uniformly spaced and
because of limited computational power and memory

Target detection and tracking
• Application of WSNs to intrusion detection, sorting and target tracking
have relatively high importance in computer and communication systems
• Generally, target detection and tracking is a well recognized discipline.
• A number of solutions in the literature exist which make use of either
domain-specific sensors or simple inexpensive tools and techniques
• Latter approaches are attractive for cases where target detection and
tracking need scale appropriately.
• Tracking with inexpensive WSN systems has certain challenges such as:
 Digital signal processing and synchronized decision making,
 Multi-modal sensing,
 High-frequency sampling and data synthesis

• Among the main applications of WSNs are:
• Recent developments in the efﬁciency of sensing, computing, and
communications technology have made it possible to use a group of
sensors inside a sensor node (sensor network processor)
 Intrusion detection
 Target detection
 Classiﬁcation
 Tracking
 Localizing

• Low cost of sensor nodes have made their deployment possible in disperse
geographic area making such devices as capable contenders for tackling
distributed detection, categorization and tracking problems
• Different type of centralized and distributed techniques have been
developed to address the mentioned problems which can be:
 Based on high message complexity, or be high computational
 Data fusion-based or decision fusion-based

• Classification problem: In classification, the target has to be identified as
vehicles, soldiers, trees, and animals.
• Classification depend on on approximation where the relevant parameters
of the sensed signal are estimated i.e. power density, period, duration,
amplitude, phase, and bandwidth.
• Main performance measures of the classification process are:
a. Probability of correct decision i.e. to correctly identifying the i-th class
b. Probability of error i.e. incorrectly identifying the i-th class as the j-th class.

• In tracking, the target position is tracked as it moves in the area covered by
the WSN.
• For a successful tracking process, the estimates of the target’s primary
point of entry and recent position must be within acceptable detection
latency.
• Some key factor that can affect the design and development of a WSN for
target detection are:
a. Energy
b. Dependability
c. Complexity

1. Energy:
 Nodes in a WSN employ either stored energy-based batteries or harvested-based
energy solar cells.
 The rate at which energy is consumed is limited either by the node’s required lifetime
for stored energy or by the mean rate of energy gathered using harvesting.
 Four major operational modes in which nodes use energy include:
 Each of these operations require speciﬁc amount of energy depending on the
effective work that is to be done.
1. Sensing the environment
2. Performing computation
3. Saving data
4. Communicating with peripherals, other nodes and interfaces.

2. Dependability:
3. Complexity:
 The instability of WSNs seriously impact the system design for classification and
tracking, especially when choosing feature that provides the basis of classification
 Two main techniques while working on feature selection:
 Complexity is associated with each operation performed by a sensor node i.e.,
detection, estimation, classification, tracking, synchronization, and routing perform
the sensing, computing, storage, and communications operations
 Therefore, the designer should concentrate on optimizing the time, space, and
message complexity of the algorithms and protocols
 Centralized
 Distributed

• Recognition of target perturbations to environment (phenomenology):
 Phenomenology is the study of the basic nature of things.
 The aim is to ﬁnd a set of characteristics with similar significance-level for objects in
the same classes and extremely different for objects in dissimilar classes.
 Features are identiﬁed as belonging to one of the six main energy domains:
 Noting that the sensors may possibly detect various facets of the same energy realm,
designers need to consider all six
1. Electrical
2. Mechanical
3. Magnetic
4. Optical
5. Thermal
6. Chemical

• Recognition of target perturbations to environment (phenomenology):
 Example:
• An automobile may disturb the surroundings electrically, magnetically, thermally, optically,
seismically, acoustically, and chemically.
• Generally, automobiles have a thermal signature consisting of hotspots like the engine area.
• In addition, automobiles have evident seismic and acoustic signatures.
• Automobiles generate carbon monoxide and carbon dioxide as a side effect of burning fuel.
• They also absorb, reﬂect, diffract, and scatter electromagnetic, optical, ultrasonic, and
acoustic signals

• Sensing selection
 Decide the correct set of sensors for the system under design so as the solution is
cost-effective and the system lifetime is maximized.
 Sensors are used to monitor variety of physical quantities e.g., electrical or optical
signature and the output of sensor is an estimate that lacks precision
 There are two main types of sensors:
• Active sensors: Can gauge a target’s existence or velocity by how the target changes, or
reﬂects a signal sent by the sensor
• Passive sensors: To identify and gauge various analogs of a target such as its acoustic,
magnetic, or thermal mark.

• Sensing selection
 Together, the speciﬁcation of the target and tracking, its design concerns, target
phenomenology, and sensing mode and domain are used to select an appropriate
sensor.
 The guidelines to be considered include the following:
1. No need for a special hardware.
2. The sensor can operate regardless of its orientation.
3. The signal processing schemes needed for related signal detection and parameter
estimation are realistic given the limitations of the situation.
4. The sensors are properly picked out and can be found by using off-the-shelf commercial
sources.
5. Co-location of sensors will not create interference or crosstalk.

• Technological advancement in WSNs have led to new applications such as
weather monitoring, smart homes and cities, health care applications,
military applications and infrastructure protection.
• Problem associated with emerging WSN applications is data aggregation
• Two main schemes are used to prevent from data aggregation problem:
Contour and edge detection
1. The ﬁrst scheme uses distributed data compression to reduce the size of data to be
transmitted.
2. The second scheme is distributed data fusion, which pre-handles the gathered data in
order to obtain synthesis results for broadcasting.
• Because of limited spectrum, high communication cost, and limited processor power
in WSNs, data aggregation from sensors pose a challenge

• Applications like edge detection produce only a binary result of 1 or 0
(indicating ) based on running the algorithm on noisy data.
• For some real-world applications, the fundamental physical events are
characterized by a range of values (not binary values)
• When applying edge detection, we digitize measured values while some
aspects of the monitored event vanish because of the quantization
• Contour lines offer additional information on the event being observed in
terms of spatial dispersal and a processed set of values of collected data
• Contour line removal has applications in the environment and in weather
forecasting, among others.

• Liao et al. devised a distributed algorithm for contour line extraction with
wireless sensor networks [10].
• The scheme consists of three main steps:
1. Consecutive extremum search: The gradient is approximated and used to find the
extremum (either the maxima or the minima) point of the uni-modal field which is
referred to as the reference point of the sensor field.
2. Sensor clustering and contour point finding: Depending on the location of the
extremum, sensors are grouped into a number of regions and clusters according to
their relative location w. r. t. the reference point. Next, contour points using
collected data samples in each cluster are determined.
3. Contour line creation: The information of contour points is shared with the class
leaders (heads) and then to the base station for the contour line creation.

• Environmental applications:
Types of applications
 Wireless sensor networks can be used effectively to monitor the environment
 Sensors are deployed to monitor different environmental factors and conditions
 Under this application category, we can identify the following cases:
1. Sensors can be deployed to observe animals or plants in the wild and monitor wild habitats as
well as the environmental factors of these habitats.
2. Water or air quality control: WSNs are used to monitor water or air quality by deploying sensors
on the ground or under water.
3. Disaster monitoring: Sensors can be deployed in the forests to detect fires or in rivers to detect
floods. In addition, seismic sensors can be deployed inside a building to find out path and degree
of earthquakes and offer an estimation of the safety of building

• Health care applications:
 WSNs can be employed to observe and trace patients and senior citizens for health
care reasons
 Possess a great potential to reduce the overall cost of health care
 Sensors can be installed in patients’ homes to monitor their behavior or movement;
vital signs of patients along with GPS coordinates can be monitored.
 Achieved by using wearable sensors that can be incorporated into wireless body
area networks (WBANs) for real-time monitoring

• Manufacturing process control:
 Sensors can be used to monitor the manufacturing processes.
 Example: small sensors can be set into the regions of speciﬁc machines and devices
that are unattainable to humans in order to monitor the condition of a machine and
send warning signal in case of any failure.
 WSNs can help decrease the cost of maintenance in factories, as well as enhance
the life of equipment and even save human lives.

• Intelligent and smart home:
 WSNs are used to provide comfortable and smart living environments e.g., sensors
can be installed in homes and connected to make an independent home network
 An intelligent refrigerator linked up to a smart microwave oven or stove can arrange
a menu based on the stock of refrigerator and send related cooking factors to smart
stove or microwave, which in turn can set the required temperature for cooking
 Utility companies use WSNs to read gas, water, and electricity meters remotely
without the need to physically visit homes to read the meters.

• Agriculture:
 WSNs are being used more and more in the agricultural industry.
 Gravityfed water systems can be observed via pressure transmitters in order to check
water tank levels.
 Irrigation automation can help in better water consumption and help reducing the
water wastage.
 WSNs can be used to control temperature and humidity levels in greenhouses so
that, when they fall below threshold levels, host systems can trigger an appropriate
response in order to ﬁx the problem.

• Military applications:
 WSNs can be used for:
• Military situation awareness
• Detection of enemy unit’s movements on land or sea
• Chemical or biological threats
• Offering logistics in urban warfare
• Battleﬁeld surveillance command, control, communications,
• Reconnaissance, computing, intelligence, surveillance, and targeting systems

 Intelligent assistance: Wireless sensors can be installed on unmanned robotic vans,
tanks, ﬁghter planes, missiles, or submarines to direct them to their goals as well as
to enable them to coordinate and collaborate with each other effectively
 Object protection: Wireless sensors can be mounted around sensitive objects such as
tactical bridges, telecommunication centers, electrical power generation stations, oil
pipelines, water puriﬁcation plants, and military nerve centers.

 Battlefield observation: Wireless sensors can be mounted in battleﬁelds to monitor
and track vehicles, and force movement, all of which permit close surveillance of
enemy forces
 Remote sensing: Wireless sensors are installed for remote sensing of biological,
nuclear, and chemical weapons, as well as for uncovering possible terrorist attacks.

• Underwater applications:
 Wireless sensors can be deployed at the bottom of oceans, seas, rivers, and lakes in
order to monitor underwater events and report them in a timely manner
 Ocean-bottom sensor nodes serve excellent means for facilitating applications for
oceanographic data gathering, contamination observation, offshore investigation,
calamity avoidance, and assisted route-ﬁnding applications.
 Unattended or independent underwater vehicles (UVs) with underwater sensors ﬁnd
applications in discovery of undersea minerals resources

 For underwater applications, wireless underwater acoustic sensor networks are used
 Underwater acoustic sensor networks consist of a number of sensors and vehicles set
up to act in a collaborative manner to monitor related events over a given area.
 By observing the underwater events, the sensors and vehicles self-organize in an
independent network that can adjust to the attributes of the ocean surroundings

 Conducting underwater search processes is challenging due to the characteristics of
water that include:
 These factors inﬂuence underwater wave propagation.
• Varying levels of salinity or measure of dissolved salts in sea water,
• Temperature
• Pressure under the surface of water.

 Traditional method for ocean-bottom or ocean-column monitoring is to install
underwater sensors that record data during the monitoring mission, and then pick up
the instruments.
 This monitoring tactic has the following drawbacks:
1. Real-time monitoring is not feasible.
2. The amount of data recorded throughout the monitoring mission by each sensor is restricted by
the capability of the on-board memory devices
3. No communication is possible between onshore control systems and the monitoring devices.
4. If a malfunction happens, it may not be possible to notice it before the instruments are
recovered
varying levels of salinity or measure of dissolved salts in
sea water, temperature, and pressure under the surface of
water.

 An underwater acoustic network (UAN) is a small-scale network deployed under
water able to gather data using acoustic modes.
 Major research challenges and design issues in the ﬁeld of UWSN are:
• Trafﬁc congestion control;
• Security, resilience and robustness
• Dependable data transfer;
• Effective multi-hop acoustic routing;
• Distributed localization and time synchronization;
• Resourceful multiple access;
• Acoustic physical layer.

 Factors influencing the underwater communication:
1. Underwater currents: Changeable flow or current speed influences the relative position of
sensors and also affect the communication quality because of the noise.
2. Salinity: Increased level of salt in the water increases the density of water which produces delay
in the signal path.
3. Multipath delay spread: Because of the various reflections of transmitted signals, the signal is
received at diverse times, which results in inter-symbol interference.
4. Surrounding noise: Such noise is generated by underwater currents, marine life and shipping in
the harbor.
5. Temperature instability: Experiments have revealed that tidally driven temperature
fluctuations have consequences on acoustic communication.
6. Multipath fading: Whenever we have waves of multipaths that are out of phase, the signal
strength at the receiver is decreased. Usually called Rayleigh fading.
7. Pressure: It has been found that high pressure at the bottom of the ocean/sea can affect the
signal communication.

 Main limiting factors in UWSN: Restricted memory
• Since sensors have limited memory, there is a requirement to reduce the data
saving and communication.
• Table-driven routing techniques with additional memory can be considered as an
option.
• Since retransmissions necessitate store-and-forward means for data packets, and
if these retransmissions are reduced, then the issue of small memory will be
lightened

 Main limiting factors in UWSN: Propagation delay
• The delay in underwater networks is relatively high compared to air, decreasing
the speed of propagation to about 1.5 × 103.
• All the propagation-delay-connected calculations during implementation should
be done using speed of sound in water.

 Main limiting factors in UWSN: Low frequency
• Communication between nodes in UWSNs takes place in very low (acoustic)
frequencies and not radio frequencies.
• Absorption attenuates the signal which is greater for higher frequencies.
• Therefore, an acoustic channel with frequency in the range 20 Hz–20 kHz is
favored.
• Hence, slighter attenuation can be obtained up to a certain level at the expense of
the propagation time.

 Main limiting factors in UWSN: Large transmission power
• Due to the resistance of water to wave propagation, attenuation is more severe
in case of UWSN
• The transmitter must send signals with higher power in order to receive a signal
with good signal-to-noise ratio.

 Main limiting factors in UWSN: Modest bandwidth
• The bandwidth for underwater acoustic channel is limited and mainly relies on
both frequency and transmission distance.
• Reason for limited bandwidth is the underwater absorption with most acoustic
systems operating below 30 kHz.
• The packets size that sensor nodes make use of must be as short as possible to
avoid high transmission delay

 To gain an insight into the communication among different sensors in a UWSN is
given in Figure below:
 The figure depict a small network seven sensors
 The sensors communicate with other with the help
of acoustic links
 Sensor 1 is communicating with sensor 2
 Sensors 4 and 7 are communicating with each
other with the help of sensors 5 and 6
 Sensor 3 is isolated since no other sensors is in its
transmission range

 Figure below depicts the architecture from the horizontal viewpoint at the sea plane.
 The ﬁgure shows the location of local sinks
and sensors at the surface of the sea.
 The local sinks are assumed to be static and
arranged in the form of a grid.
 This way, the local sinks are capable of
communicating with each other via both
acoustic and radio frequency forms

 This figure shows another
architecture from the angle of the
vertical view at the sea surface
 The relative place of the sensors is
in the water column under the sea
surface.

Module 3
Medium Access Control (MAC) in Wireless Networks

Contents
1. Medium access control (MAC) in wireless networks
2. MAC layer protocols in WSNs and their performances
3. MAC issues and challenges in WSNs

1. Discuss the medium access control (MAC) in wireless
networks.
2. Describe the various MAC layer protocols in WSNs and their
performances.
3. Understand various MAC issues in WSNs.

Required Reading
1. Chapter 4: Principles of Wireless Sensor Networks, M.
Obaidat and S. Misra, Cambridge University Press, 2014.
(ISBN: 978-0-521-19247-7)
Recommended Reading
1. Chapter 5: Wireless Sensor Networks, Ian F. Akyildiz and Mehmet Can
Vuran, John Wiley & Sons, 2010. (ISBN: 978-0-470-03601-3)
This Presentation is mainly dependent on the textbook: Principles of Wireless Sensor Networks, M. Obaidat and S. Misra,
Cambridge University Press, 2014. (ISBN: 978-0-521-19247-7)

• Medium access control in wireless networks

Medium access control in wireless networks
 Standard protocols in wired local area networks  Carrier Sense
Multiple Access with Collision Detection (CSMA/CD) scheme to access
the medium.
IEEE 802.3 Ethernet.
Slotted ALOHA networks -: use a Time Division Multiple Access
(TDMA) scheme.
CSMA/CD based Ethernet : better performance in Low node density
network

MAC in WSN
Primary goal : Energy Saving
Factors consuming energy :
Transceiver devices
Restricted range of transmission
Characteristics of MAC for WSN : Energy Efficiency

• Sources of Energy consumption
• Transmission of overhead
• Transmission of overhead caused due to exchange of data based on the size of the
packet.
• Idle listening
• Switch of the device battery when in idle mode for a specific duration in order to save
their non-renewable source of energy power
• Localization
• Energy is consumed when the nodes identify their placement coordinates during
mobility.

• Auto configurable networks
• Energy is exhausted when the nodes are self organizing due to failure of any nodes in the
network.
• Collaborative function
• Increasing the efficient data collection and aggregation from different nodes in WSN is a
source of energy consumption.

MAC layer protocols in WSNs and their performances
• Major MAC Layer Protocols for WSN
S-MAC – Sensor MAC
L-MAC – Lightweight MAC
D-MAC – Dynamic Scheduling MAC
Energy-efficient QoS-aware medium access protocol.
Energy-efficient application-aware medium access protocol.
A location-aware access control protocol.
An energy-efficient MAC approach for mobile WSNs.
O-MAC: A receiver-centric power management protocol.
PMAC: An adaptive energy-efficient MAC protocol for WSNs.
T-MAC protocol: Timeout-MAC
BMAC protocol: Berkeley MAC

• S-MAC: An energy-efficient protocol
To increase energy conservation and support self-configuration.
• Identifies two main energy consuming mechanisms.
Collision- Retransmission of corrupted data packets due to collision
Control packet overhead- Sending and resending of control packet
Overhearing- a node keeps listening to the data packets that are not intended
for itself.
Idle Listening : 50% - 100% energy consumption due to listening to the other
nodes periodically

• S-MAC
Reduce the consumption of energy from all the above sources.
Lets nodes periodically enter into the sleep cycle or idle mode.
The PAMAS (power aware multiple access protocol) uses the same
method to avoid idle listening, which leads to the reduction of energy
consumption.

• Network and applications assumptions-
WSNs are in many ways different from traditional IP networks.
AWSN consists of many tiny devices that are separated by a very small
distance from each other and from the target.

• Periodic listening and sleep
Nodes are assumed to have a sleep cycle
period. See Figure 1.
In a cycle of one second, a node listens to
the medium for half a second and sleeps
for the other half;
this will lead to a 50% reduction in the duty
cycle.
This implies 50% saving in the energy.
• Basic scheme
The node goes into periodic sleep cycles if
there is nothing to listen to on the channel.
It then wakes up to see if any other node
wants to talk to it.
Nodes exchange their schedules by
broadcasting them to their neighbours; see
Figure 2.
Figure 1: The division of time into listen and
sleep states in the S-MAC protocol.
Figure 2: Illustration of two nodes, A and B, having
different schedules; thus they synchronize with nodes C
and D respectively.

• Selecting and maintaining
schedules
Every node maintains a table called
the scheduling table.
• Maintaining synchronization
Update the schedules to bring
synchronization between the
elements.
Periodically sending the SYNC
packet See Figure 3
Figure 3: Illustration of different sender configurations.

• Collision and overhearing avoidance
The protocol uses the RTS/CTS mechanism of the IEEE 802.11 protocol
• Performance characteristics of S-MAC
S-MAC uses mechanisms
• Collision avoidance.
• overhearing prevention.
• less idle listening.

• L-MAC: a light-weight medium access protocol
Light-weight energy-efficient scheme
Minimize the overhead at the physical layer.
Decrease the states of the transceiver switches
Minimize energy wasted in the preamble transmission.
Save power by making the sleep interval of the node adjust to the volume of traffic.
This mechanism reduces the complexity at the physical layer.

• L-MAC: a light-weight medium
access protocol
L-MAC protocol characteristics
1) The transceiver is assumed to
have a single channel
2) Has three states of operation:
receive, transmit and standby.
3) A transmission mode uses more
energy than the receive mode.
The control field is again divided into
different parts as shown in Table 1.
Table 1: Different fields in the control field of the L-MAC protocol

• Network setup
All nodes are unsynchronized.
The gateways take the initiative and start controlling slots.
The immediate one-hop neighbours will start synchronizing.
The recently synchronized nodes will pick up random time slots
All nodes are required to maintain a lookup table.

• Performance characteristics
MAC protocol performs well against the S-MAC and the very old EMAC
protocols.
This lightweight nature comes from the time being divided into equal slots.
Since there are no contentions, the nodes do not waste their energy in
retransmitting lost or damaged messages.

• Dynamic scheduling MAC protocol
To designed for the requirements of very-large-scale sensor networks.
Adds a dynamic scheduling scheme to the modified distributed mediation
device (MDMD) protocol.
Before designing any new MAC protocol for WSNs, the following points need
to be taken into consideration.
1) Energy efficiency.
2) Scalability.
3) Adaptivity to changes in node density and topology.

The dynamic scheduling MAC
protocol is founded on the device
mediation device (DMD) protocol.
The DMD protocol brings in a new
kind of node called the mediation
device (MD) node.
Such a node acts as an
arbitrator/mediator between the
other wireless sensor nodes; see
Figure 4.
Figure 4: Illustration of RTS–CTS exchange in a dynamic scheduling
MAC.

• Modified distributed mediation device (MDMD)
Minimizes the delay introduced by the distributed mediation device protocol.
Each node has got its ON state followed by its OFF state.
Each node sends a query packet at the end of its ON state.
All nodes make and fill their database in the listening period.
Every entry in the database includes the information about the ID and its
timetable information.

Every time there is any fresh node joining the network.
It has to listen to the full cycle of the sleep–listen period.
Every node also updates its database periodically.
Every time a node needs to transmit data to another node there may be a
third node that can interfere with its transmission.

• Neighbour-aware dynamic (NAD) scheduling
NAD introduces a dynamic schedule scheme to resolve the hidden and
exposed terminal/node problems.
The exposed node/terminal problem happens when there are two groups of
neighbour clusters.
The transmitter nodes are in direct range of each other.

The transmitter in the other group is averted from transmitting.
Figure 5 shows an instance of this type of problem.
Figure 5: Illustration of the exposed node terminal problem between nodes S1–R1 and S2–R2.

The hidden terminal (node) problem occurs when we have more than one
transmitting node that wants to send data to the same receiver.
Such that the receiver is in the range of both transmitters, but the two
transmitters are not in the direct range of each other.
In such cases if just a DATA-ACK sending mechanism is used the data
transmission will result in a collision at the receiver.
In order to avoid such situations a common mechanism called RTS–CTS
(request-to-send–clear-to-send) communication is used [41].

Here, the transmitters first send the
receiver the RTS packet.
After receiving the RTS packet the
receiver sends back a CTS packet
only to the transmitter from whom
it received the first RTS packet.
 Then, the receiver does not
respond to any other RTS packets
from other transmitters until the
start of the data transfer.
• Figure 6 illustrates the concept.
Figure 6: Illustration of the hidden node/terminal problem

The two neighbouring nodes do not
have the same schedule.
If the schedules of the two nodes
coincide then they will send and
receive data at the same time.
As seen in Figure 7, the schedules
of nodes A, B, C, and D overlap even
though they are not direct
neighbours.
Figure 7: The hidden node problem in wireless sensor networks.

• Energy-efficient QoS-aware medium access (Q-MAC) protocol
It pays more attention to the QoS aspects of the network.
It is essential for energy efficiency and transmission dependability.
This scheme may be called the priority-based random access protocol
The Q-MAC scheme consists of both intra-node scheduling and inter-node
scheduling.

The intra-node scheduling utilizes a multi-queuing architecture to organize
data packets based on their application and MAC layer abstractions.
The inter-node scheduling algorithm deals with giving access to the channel
competing with various neighbouring nodes.
The techniques used in this case are the power-conserving MACAW and the
loose priority random access scheme.

• Intra-node scheduling
It employs multiple First In First Out (FIFO) order queues.
The received packets are ordered and stored in different queues based on the
criticality.
The service rank in the network is decided by the number of queues for each node.
The priority of an incoming packet is decided by the application and MAC layer
abstractions.
The application layer allocates priority to the packets based on significance of the
contents of the packet.

• Inter-node scheduling
To manage and schedule the data transmission between various nodes using
the same channel.
Because of the high cost connected with retransmissions, inactive listening,
collision, communication overhead and overhearing.
Q-MAC employed basic and distributed protocols such as the power
conservation–multiple access with collision avoidance for wireless (PC-
MACAW), and loosely prioritized random access (LPRA) protocol.

In a PC-MACAW protocol- each frame is described as the set of RTS–CTS–
DATA–ACK cycle.
Each frame period is associated with a small space known as frame space(See
Figure 8.)
Figure 8: The frame format in PC-MACAW protocol.

• Energy-efficient application aware medium access protocol
Energy-efficient medium access protocol.
Achieves energy efficiency by avoiding idle listening, data collision, and
transmissions to a node .
It experiences less delay and fewer collisions.
To achieve energy-efficient channel access.
Will need the two-hop neighbourhood and flow information to perform the
election process.

• As shown in Figure 9, time is arranged into the period of random and
scheduled access slots.
Figure 9: Illustration of the time slot division into different components in a QoS-aware MAC.

• Location-aware access control protocol
Cantered on roughly defined areas of several access points.
The location key is obtained from the beacon information.
An open-air medium, numerous techniques can be used.
Typically, all of these methods need some kind of password, ticket, biometric
or access control list (ACL) to verify an entity.

• Among these methods are the ones listed below.
Access control list (ACL).
• Keeps a list that contains info on what each user is entitled to access in a specific setting.
Secure ID.
• This is a token-based authentication method.
Kerberos.
• Ticket-based authentication technique in which each user is assigned a ticket for a
session.

• An energy-efficient MAC approach for mobile wireless sensor
networks
Since in wireless networks the nodes are on the move, this creates a problem
called the Doppler effect/shift, which leads to the losses.
Reduces the Doppler effect by effectively changing the frame length of the
data during transmission.
A filter identified as the extended Kalman filter is used to forecast frame size
for every transmission.

Deals with how to increase the energy efficiency for a network.
Always experiences a Doppler-shift effect.
The noise in the wireless channels also leads to a low signal-to-noise ratio.
The retransmission can lead to the loss of a large amount of the energy.
This scheme suggested an algorithm that may lead to the reduction of the
retransmission energy losses.

• O-MAC: a receiver-centric power management protocol
This is based on Pseudorandom Staggered On.
O-MAC can achieve near optimal energy efficiency.
Two variations of the O-MAC protocol – with local broadcast channel and
preamble-sized slots.

Solving issues such as time synchronization, and management of neighbour
tables is essential for the development of the O-MAC protocol.
Time synchronization- The Staggered On and Pseudorandom Staggered On
processes need time synchronization.
Asynchronous neighbour discovery- O-MAC contains an inhabitant discovery
mechanism based on load-balanced beaconing.
We have two types of nodes: synchronized and unsynchronized.

• PMAC: an adaptive energy-efficient MAC protocol for wireless sensor
networks
Instead of having pre-set sleep–wake up cycles, the duty cycles of the nodes
are found out dynamically.
This scheme can offer power savings throughout low-load situations and
enhancement of the throughput throughout high-load situations.
The pattern MAC scheme considers all the major factors for its design; namely
power efficiency, latency, and throughput.

• Figure 10 illustrates the point
that, under a no-traffic situation
S-MAC- scheme with a fixed duty
cycle.
T-MAC scheme with a variable duty
cycle run into some waste in the
power.
PMAC scheme, which adjusts itself
based on the traffic status, will
inform the nodes to sleep.
Figure 10: An illustration that compares the lengths of the idle
listening periods of S-MAC, T-MAC and
LMAC with the no-traffic condition.

• T-MAC
Contention-based MAC protocols for wireless LANs.
Introduces the idea of a variable duty cycle compared with the fixed duty
cycle of S-MAC and the no-duty cycle of CSMA.
 Demonstrates an enormous amount of performance increase when
compared to the S-MAC protocol.

Each node regularly awakens in a specified time frame
Begins receiving and transmitting data if there are any, then returns to
sleep.
The activation episode here indicates the following:
(a) the event of the notice of the recurring timer.
(b) the event of the delivery of data via the channel.
(c) the event of perceiving the communication.
(d) the event that shows data swap of the neighbour has finished.

• BMAC protocol
It is a configurable protocol that is simple in its design and implementation,
but is efficient.
Provides a good interface for WSNs consuming low power.
Very efficient in collision avoidance and has high channel utilization.
Provides a preamble sampling mechanism to minimize the duty cycle and
decrease idle listening.

• BMAC goals
a) small power operation
b) implementation must be as simple as possible
c) channel utilization should be as optimal as possible at both high and low
data rates
d) it must have an effective collision avoidance technique
e) it should be scalable with the increasing network.

BMAC technique is a very small core medium access protocol.
It does not offer any type of network layer organization or
synchronization strategies.
BMAC uses a method called
• Clear channel assessment and packet back offs for channel arbitration
• Low-power listening (LPL) for lower power consumption
• Link layer acknowledgements for reliability.

BMAC does not offer any multi-packet methodology such as RTS/CTS
exchange.
BMAC often offers the link layer acknowledgement support.
The BMAC also utilizes a scheme of low power listening.

• MAC issues in wireless sensor networks

MAC issues in wireless sensor networks
1) A WSN consists of a much larger number of nodes than traditional
wireless networks.
2) The WSN topology changes more frequently due to high probability
of node failure and mobility.
3) Nodes of a WSN have limited computational power and storage.
4) Nodes in a WSN are typically powered by batteries, hence they have
limited life.
5) Nodes in a WSN are usually set up in an ad-hoc manner, thus they
should arrange themselves into a communication network.

Main Reference
1. Chapter 4 (Principles of Wireless Sensor Networks, M.
(ISBN: 978-0-521-19247-7)

Module 4
Routing in Wireless Sensor Networks

Contents
1. Fundamentals of routing and challenges in WSNs.
2. Network architecture-based routing protocols for WSNs.
3. WSN routing protocols based on the nature of operation.

1. Understand the basics of routing and related challenges in
WSNs.
2. Explain the various routing protocols in WSNs.

Required Reading
1. Chapter 5: Principles of Wireless Sensor Networks, M.
(ISBN: 978-0-521-19247-7)
Recommended Reading
1. Chapter 7: Wireless Sensor Networks, Ian F. Akyildiz and Mehmet Can
Vuran, John Wiley & Sons, 2010. (ISBN: 978-0-470-03601-3)

• Routing in wireless sensor networks

Fundamentals of routing and challenges in WSNs.
• The major design issues of routing in WSNs are
1) Power spending without affecting correctness.
2) Node setup.
3) Data-sending paradigm.
4) Fault tolerance.
5) Node dissimilarity.
6) Scalability.
7) System dynamics.
8) Transmission media.
9) Connectivity.
10) Quality-of-service (QoS).
11) Data aggregation
12) Coverage.

Wireless sensor networks may
contain numerous nodes
These wireless sensors can have the
ability to communicate among each
other.
The great number of wireless
sensors can allow the sensing of the
network
Figure 1 shows a schematic diagram
of the components of a WSN.
Figure 1: The main components of wireless sensor networks (WSNs).

Each of the wireless sensor nodes
has the ability of sensing,
processing, and transmission.
Limitations
• Limited computation power
• Limited memory
• Limited power supply
• Limited bandwidth and data rate (see
Figure 2).
Figure 2: A typical wireless sensor network (WSN).

Routing Protocols in WSN – A Taxanomy
Ref : Jamal N. Al-Karaki, “Routing Techniques in Wireless Sensor Networks: A
Survey”,

• Network architecture-based routing protocols for wireless
sensor networks (WSNs)

Network architecture-based routing protocols for WSNs
• Routing in WSNs can be divided into three categories:
a) Flat-based routing
b) Hierarchical-based routing
c) Location-based routing depending on organization of the WSN.
• A routing protocol is called adaptive if specific parameters can be managed
in order to adjust the current network conditions.
• Such protocols can be categorized into:
• Multipath-based
• Query-based
• Negotiation-based
• QoS-based
• Coherent-based routing techniques.

• Routing schemes can divide into:
• Proactive
• Reactive
• Hybrid
Based on how the source discovers a route to the target.
In proactive schemes, all routes are calculated ahead of the time
when they are needed.
In reactive schemes, routes are calculated when needed.
The hybrid protocols use a blend of these two schemes.

• Multi-hop flat routing
Each sensor node performs similar tasks to the others.
The nodes cooperate together to do the task.
Here, it is not realistic to give a global ID to each sensor node.
This situation has led to data-centric routing, in which the base station (BS)
transmits inquiries to specific areas and waits for the data from the sensors
located in there.

• Sensor protocols for information via negotiation (SPIN)
It spreads the information from every sensor node to all other nodes.
Makes use of the feature of clustering where nodes close to each other have
similar data.
Data negotiation and resource-adaptive algorithms are often used.
The main concepts on which the design of SPIN schemes is based are
a) Nodes function well and save energy by sending some of the data,
b) Traditional schemes such as flooding or gossiping-based routing protocols.

Since the nodes use three kinds of messages
ADV, REQ and DATA – to correspond, SPIN is considered a three-stage scheme.
There are many protocols that have been devised under the SPIN family of
protocols.
These include SPIN-1 and SPIN-2, which involve negotiation before sending
data to guarantee that only needed information will be sent.
There are other protocols in the SPIN family.
 SPIN-BC
 SPIN-PP
 SPIN-EC
 SPIN-RL.

• Directed diffusion
This was devised as a data-aggregation scheme.
It is a data-centric and application-aware model.
Data are combined from different sources while traveling by getting rid of the
redundancy
Reducing the number of transmissions
Hence saving the power of nodes, and extending the overall network lifespan.

• Directed diffusion
This scheme finds routes from multiple sources that are intended for the
same destination.
In this protocol, sensor nodes assess events and build gradients of
information in their neighbourhoods.
The base station asks for data by sending inquiry messages describing a
specific task to be done by the WSN.

• Minimum cost forwarding algorithm (MCFA)
The MCFA scheme [18] utilizes the situation that the path of routing is at all
times known, which is to the base station.
Therefore, a node does not need to have a distinctive ID nor keep a routing
table.
Each sensor node keeps the smallest cost estimate from itself to the base
station.

• Minimum cost forwarding algorithm (MCFA)
All messages that have to be sent by the sensor node are relayed to its
neighbours.
If a node gets the message, it tests if the message is on the cheapest cost
path connecting the source node and the base station.
If it is true, then it re-sends the message to the nearby nodes.

• ACQUIRE protocol
Based on querying the WSN.
It perceives the network as a distributed database with compound queries.
The BS node transmits a query that is then advanced by each node getting
the query.
Each node attempts to reply to the query in part by utilizing its pre-cached
information and then moves it to another node.

• Energy-aware routing protocol
To extend the lifetime of the WSN by saving consumed power as much as
possible.
It keeps a set of paths instead of keeping one best possible path at high data
rates.
Such paths are preserved and selected using a specific probability function.

• Rumour routing
It is a variation of the directed diffusion scheme.
It is used for situations where geographic routing is not possible.
It employs flooding to insert the query to the whole WSN.
It has been reported that rumour routing provides considerable energy
savings if compared to event flooding.

• Gradient-based routing
This protocol is a variant of the direct diffusion scheme.
Learn the number of hops when the attention is diffused throughout the
entire WSN.
Every sensor node is able to compute a parameter termed the height of the
node.
This approach uses some ancillary techniques such as data accumulation and
traffic distribution .

• Gradient-based routing
It has three data dissemination schemes-
• A stochastic technique
• Where a node chooses arbitrarily one gradient if two or more subsequent hops have
similar gradient.
• An energy-based technique
• Where a node enhances its altitude if its energy goes under a specific edge.
• A stream-based technique
• Where new streams are not sent over the nodes that presently belong to the paths of
additional streams.

• Routing protocols with random walks
This family of protocols obtains load balancing in a statistical manner.
It is meant only for giant networks with partial movement for the nodes.
Every sensor node has a distinctive identifier.
In order to discover a route from a source to endpoint, the position
information is acquired by calculating distances among nodes by the
distributed asynchronous form of the Bellman–Ford routing protocol.

• Information-driven sensor querying (IDSQ) and constrained
anisotropic diffusion routing (CADR)
In IDSQ, the probing node can decide which node can deliver the most
valuable information with the extra benefit of equalizing the power bill.
IDSQ does not precisely describe how the request and the data are
transmitted among sensor nodes and the BS.
CADR is meant to be a general form of directed diffusion protocol.
The major concept here is to probe sensor nodes and move data in the
network.

• COUGAR
This protocol is considered a data-centric protocol, which regards the WSN as
a giant distributed database structure.
It employs declarative queries so as to extract query handling from the
network layer tasks like selection of applicable sensors.
It includes a structure for the sensor database system in which nodes choose
a head node in order to achieve aggregation and send data to the base station
(BS).

• Hierarchical/cluster-based routing schemes
Originally devised for fixed communications networks.
Their key advantages are scalability and efficient communication
characteristics.
Utilized to accomplish energy-efficient routing in WSNs.
Higher-energy nodes are able to process and send the information whereas
low-energy nodes can carry out the sensing in the vicinity of the target.

• LEACH (low-energy adaptive clustering hierarchy)
Designed for WSN that require an end-user to remotely observe the
environment.
Cluster-based scheme, which includes distributed cluster formation.
It arbitrarily chooses a few sensor nodes as cluster heads (CHs) and alternates
this role to uniformly distribute the energy load between the sensors in the
network.
LEACH uses a TDMA/CDMA MAC scheme in order to decrease inter-cluster
and intra-cluster collisions.

• Power-efficient gathering in sensor information systems (PEGASIS)
To prolong the WSN lifetime.
Nodes should correspond only with their nearest neighbours.
Nodes should alternate in corresponding with the base station (BS).
This can minimize the power needed to send data per cycle .

• Power-efficient gathering in sensor information systems (PEGASIS)
This scheme has two main aims:
To increase the lifespan of every node by employing cooperative methods
To facilitate only neighbouring management between nodes that are near to each other.
The clustering overhead is prevented, PEGASIS necessitates dynamic topology
tuning as a sensor node
This kind of regulation may produce substantial overhead, particularly under
heavy load conditions.

• Threshold-sensitive energy-efficient protocols
There are two known hierarchical routing algorithms that fall under this category:
• Threshold-sensitive energy-efficient sensor network protocol (TEEN).
• Adaptive periodic threshold-sensitive energy-efficient sensor network protocol (APTEEN).
These schemes are recommended for time-constraint applications.
Nodes sense the medium constantly, but data transmission is rarely performed.
Each cluster head sensor transmits to its group a strict limit .

This prompts the node to turn on its transmitter and send data.
The strict limit attempts to minimize the number of transmissions.
The APTEEN scheme is a hybrid protocol, which modifies the ceiling values
used in the TEEN technique.

a) The schedule, which is a TDMA timetable that assigns a time to each node;
b) The count time, which is the largest time period between two consecutive
reports transmitted by a node;
c) Attributes, which are a group of physical values that the user has interest in
getting reports on; and
d) Limits (thresholds), which are a set of strict or soft ceilings/thresholds [38–
40].

• Small minimum energy communication network (MECN)
A particular sensor network by using low power global positioning systems
(GPS).
It identifies a relay region for each sensor node in the network.
The communication area comprises nodes in a nearby area.
This scheme is self-reconfiguring
Another efficient scheme, which is an extension of MECN, is called small
minimum energy communication network (SMECN) [3, 42].

• Self-organizing protocol (SOP)
Employed to form a structure used to sustain mixed types of sensors.
Such sensors can be transportable or fixed.
Router nodes are fixed and they establish the pillar for interaction.

• Sensor aggregates routing
Several schemes have been devised to build and support sensor collection.
The aim is to jointly observe target action in a specific setting.
A wireless sensor collection consists of those nodes in a WSN that fulfil a
predicate for a cooperative processing mission.
These factors rely on the mission and its resource constraints.

• Sensor aggregates routing
Grouped together based on their sensed signal intensity.
There is a single peak for each group/cluster.
Next, the local cluster heads are selected.
In order to choose a leader, data interchange among immediate sensors is essential.
The sensor node interchanges packets with the nearby sensor nodes
The tracking scheme supposes that the head recognizes the geographical area of the
cooperation.

• Virtual grid architecture routing scheme (VGA)
This protocol was devised as an energy-efficient routing scheme.
It employs data accumulation and in-network handling in order to extend the
network’s life-cycle.
Square clusters have been utilized to get a static rectilinear virtual topology.
In every region, a sensor node is chosen to operate as a head of the cluster.
The group of heads of clusters, which are often called local aggregators (LAs),
execute the local aggregation.

• Hierarchical power-aware routing (HPAR)
Splits the WSN into groups of sensors.
Every group of sensors is dealt with as one entity.
For routing, each sector decides how to route a message across the other sectors.
The message is sent over the path that experiences the greatest overall minimum of
the remaining power, usually called the max–min path.
The idea is that selecting the sensor nodes with the great remaining power might be
expensive.

• Hierarchical power-aware routing (HPAR)
A scheme, called the max–min zPmin algorithm.
Based on the compromise between reducing the total power expenditure .
It attempts to improve a max–min route.
Initially, the scheme discovers the path with the minimum power spending
(Pmin).
Then it discovers a link that exploits the smallest residual power in the WSN.

• Two-tier data dissemination (TTDD)
The two-tier data dissemination (TTDD) scheme offers data provision to various
mobile Base Stations (BSs).
Every source of data constructs a grid that is employed to spread data to the mobile
sinks .
In this scheme, the nodes are fixed and location aware.
When an event happens, nearby sensor nodes manipulate the signal .
In order to construct the grid, a data source announces and selects itself as the initial
crossing point of the grid, and conveys a message for its four adjacent crossing
points.

• Location-based routing schemes
In this class of routing, the nodes are addressed using their locations.
The distance between nearby nodes may be approximated based on the
power level of the received RF signal.
Comparative coordinates of adjacent sensor nodes may be acquired by
exchanging these data among nearby nodes.
In order to save power, various location-based techniques require nodes to
nap when there is no action.

• Geographic adaptive fidelity (GAF)
This is an energy-aware location-based routing scheme that was devised
mainly for mobile ad-hoc network systems.
Here, the WSN is partitioned into regions that make an implicit grid.
The nodes in the WSN can elect one of them to remain alert for a specific
period of time and then they turn to the sleep mode.
This selected node is in charge of observing and sending information to the
BS.

• The operation of this scheme is
based on three states: the
discovery, active, and sleep states.
• The GAF scheme is usually
realized for both mobile and non-
mobile environments.
• See Figure 3. Figure 3: One possible scenario of fixed zoning in WSNs.

• Geographic and energy-aware routing (GEAR)
Limits the number of notices in a straight transmission;
Having a specific region instead of transferring the request to the entire WSN.
Every sensor node in GEAR maintains an approximated cost and a learned
cost of reaching the target node via its neighbours.
The learned cost is an enhancement of the approximated cost that takes care
of the routing near holes in the WSN.

• MFR (most forward within radius), GEDIR (the geographic distance
routing), and DIR (compass routing method referred to in the
literature as DIR)
These three schemes are based on the concepts of basic distance, progress,
and direction.
The main concern here is how to handle the forward direction and backward
direction.
The GEDIR scheme is of a greedy nature, as at all times it routes the packet to
the node.

• The greedy other adaptive face routing (GOAFR)
This scheme usually selects the neighbour nearest to a sensor node to be the
following node for transmitting.
It may not find any closer node except the current node.
The other face routing (OFR) scheme is an alternative of the familiar Face
Routing (FR) algorithms.
The latter scheme is the earliest technique that ensures realization if the
sender and the receiver are linked.

• WSN routing protocols based on the nature of
operation

WSN routing protocols based on the nature of operation
• Query-based routing approach
The target nodes transmit a request for data from a node via the WSN.
Every time a mediator/agent traverses a route with a path heading to an
occurrence that it has not yet met.
Tt produces a route state that heads to this occurrence or event.
In general, a node will not produce a request unless it acquires a path to the
necessary event.

• Multipath routing schemes
Here multiple paths are employed rather than only one route. This is done to
improve the WSN performance.
A substitute route occurs between a sender and a target when the key route
dies.
This may be improved by providing several routes between the sender and
the target.

• Multipath routing schemes
Such substitute routes remain active via transmitting regular notices.
WSN stability may be enhanced by raising the communication overhead
needed to preserve the other routes.
A scheme has been devised to move data over the route with nodes having
the greatest remaining power.

• Coherent and non-coherent processing
There are two cases of data processing operations that were meant to be
implemented in wireless sensor network systems:
• Coherent data-processing-based routing
• Non-coherent data-processing-based routing.
In the latter, sensor nodes can locally treat the basic data before being
transmitted to other sensor nodes in the network for more treatment.
In the coherent routing technique, data are sent to combiners once some
minor treatment is performed.

• Quality-of-service (QoS)-based routing schemes
Designed to have a sense of balance between QoS and power spending,
The sequential assignment routing (SAR) protocol was devised.
The earliest routing scheme for wireless sensor networks to present the
concept of QoS in routing.
In this latter scheme, routing decisions are based on the following aspects:
QoS on every route, energy supply, and precedence degree of the packet.

Hence, the SAR scheme is considered a table-driven multi-route technique.
There is another scheme that comes under the QoS routing protocols; this is
called the SPEED scheme.
It offers smooth real-time end-to-end QoS assurance.
It requires every sensor node to keep data on its adjacent nodes and utilizes
geographic transmitting in order to locate the routes.

• Negotiation-based routing schemes
These protocols employ high-level data description so as to remove the
unnecessary data communication via conciliation.
The decisions on communication are based on the existing resources.
The key drive here is that the utilization of flooding to broadcast data may
yield overlap between the transmitted data.

Main Reference
(ISBN: 978-0-521-19247-7)

Module 5
Transport Protocols for Wireless Sensor Networks

Contents
1. Fundamentals of routing and challenges in WSNs.
2. Network architecture-based routing protocols for WSNs
3. WSN routing protocols based on the nature of operation

1. Understand the basic requirements of transport protocols for
WSNs.
2. Discuss the suitability of Internet transport protocols for WSNs.
3. Discuss existing transport protocols for WSN.

References
Chapter 6
(ISBN: 978-0-521-19247-7)

Introduction
• In WSN, data exchanged between the source sensor nodes and the sink
node passes through multiple hops where each hop represents a different
sensor node.
• One of the characteristic
of the network trafﬁc is
its funnel-like structure
between the source
nodes and the sink node

Transport protocol requirements for WSNs
• Transport layer protocols support two main functions:.
• Reliable data transmission in WSN environment is a challenging task due to
the following reasons:
i. Sensors have limited computation and communication power, i.e. a sensor has
limited processing power and short communication range.
ii. Sensors are battery powered. So, energy conservation is an important issue.
iii. Sensors are deployed close to the ground and this increases the unreliability of the
communication channel due to signal attenuation, channel fading or shadowing.
iv. Dense deployment of sensors increases channel contention and congestion.
1. Congestion control
2. Reliable data delivery or recovery from packet loss

Performance metrics
• End-to-end reliable event transfer and event reliability
 End-to-end reliable event transfer is performed when a sink receives the ﬁrst message
that reports an event.
 If m is the ﬁrst message that reports event e, then the probability of successful event
transfer can be expressed as:
N = set of nodes that detect event e
= link state between node si and s0
 If E events occur within an update interval, then event reliability, which is the ratio of
successfully delivered messages, can be expressed as
where

Performance metrics
• Node reliability
 Node reliability of a node i can be defined as
• Congestion detection
 Congestion degree specifies the current congestion level at each sensor node.
 Congestion degree can be defined as:
where, ti
s denote average packet service time and ti
a is the average packet interarrival
time over a predefined time interval at sensor node i

Performance metrics
• Network efficiency
 Network efficiency η is defined in as the ratio of the total number of hops traveled by
useful packets to the total number of packet transmissions in a network.
 A useful packet can be defined as a packet, which is ultimately delivered to the sink.
 The network efficiency is expressed by equation
where UP denotes the set of useful packets and P is the set of all transmitted packets,
hops(p) denotes every hop taken by p, and Txs(p,h) expresses the number of
transmissions taken by p at each hop.
 All the retransmissions and the transmissions of dropped or corrupted packets are also
taken into account in the total number of packet transmissions.

Performance metrics
• Node efficiency or imbalance
 Node efﬁciency or imbalance ζ is used to measure the performance of each node.
 Imbalance of node i can be calculated as

Performance metrics
• Node efficiency or imbalance
 Network fairness implies how, fairly or equally, each node of the sensor network gets
the chance to use the network resources and to transmit its data.
 Network fairness ϕ(i) of a node i can be calculated as
where N = total number of nodes
ri = average packet delivery rate of node i

Internet transport protocols and their suitability for use in WSNs
• In Internet, the intermediate nodes act
as layer-3 devices (layer-3 switches or
routers).
• In sensor networks, all nodes possess
transport layer. (See figure)

Internet transport protocols and their suitability for use in WSNs
• The most commonly used reliable transport protocol designed for the
Internet is transmission control protocol (TCP).
• With reference to WSNs, some of the functional properties TCP make it
unsuitable for use:
1. Typically, an end-to-end connectivity between the source and the sink nodes is not
established during a communication session In WSNs.
2. WSNs are characterized by high degrees of error-proneness associated with poor
channel quality, low bandwidth, frequent failure of sensor nodes, and congestion
3. Each intermediate node can store the data packets for a longer period of time and
then send them to the next hop, as the channel is available.
4. Resource limitations of the nodes in WSNs.

Existing transport protocols for WSNs
• Transport layer protocols for WSNs can be classiﬁed into two types:
• Only a few of these
protocols provide both
congestion control and
reliability
 Congestion control protocols
 Reliability protocols.

Congestion and flow control-centric protocols
1. Congestion detection and avoidance (CODA):
 In CODA, congestion is detected on the basis of the queue length of packets at the
intermediate nodes.
 As shown in figure, CODA comprises of three mechanisms:
 Controls the rate of ﬂow of packets based on the additive increase and multiplicative
decrease (AIMD) algorithm.
 AIMD is an energy efﬁcient technique, but the successful delivery of packets to the
destination is not guaranteed.
1. Congestion detection
2. Open-loop hop-by-hop backpressure
3. Closed-loop multi-source regulation

1. Congestion detection and avoidance (CODA):
 Congestion detection
 Open-loop hop-by-hop backpressure
 Closed-loop multi-source regulation

2. Congestion control and fairness (CCF):
 Controls congestion is based on packet service time by adjusting transmission rate
 Designed to control congestion while ensuring fairness in the delivery of packets to
the base station.
 The issue of fairness deals with ensuring that an equal number of packets is received
from each sensor node in the network over fixed time period.
 CCF has two major design considerations:
a. Congestion control design
b. Fairness design.

3. Priority-based congestion control protocol (PCCP):
 Maintains a priority index, which represents the importance of each node.
 Packet inter-arrival time and the packet service time are used to compute the degree
of congestion.
 Node priority index and the degree of congestion values further help in imposing
hop-by-hop congestion control.
 PCCP as a faster and more energy-efficient congestion control algorithm than CCF.
 Each node in PCCP is modeled to have a scheduler between the network and the
MAC layers, as shown in Figure 6.5. The scheduler is tasked to maintain two queues –
one for source traffic and the other for transit traffic.
 Weighted fair queuing (WFQ) or weighted round-robin (WRR) algorithms are to
impose fairness between the two traffic types and fairness between all sensor nodes

 PCCP consists of three mechanisms:
a. Intelligent congestion detection (ICD)
b. Implicit congestion notiﬁcation (ICN)
c. Priority-based rate adjustment (PRA)

4. Trickle:
 The key features of Trickle is the capability to limit the number of packets thus
adjusting to the packet transmission rate.
 A technique designed to propagate code updates from the downstream nodes
towards the sink node, through the intermediate nodes in the multi-hop path.
 A periodical event in each node suppresses broadcasting if the meta-data that it
receives from its neighboring node exceed the threshold.

5. Fusion:
 Controls congestion using three components that operate in a concerted manner at
the different layers of the network protocol stack:
i. Hop-by-hop ﬂow control
ii. Source rate limiting scheme
iii. Prioritized MAC layer.

6. Siphon:
 Addresses the issue of overload traffic management using the concept of multi-radio
virtual sinks (VSs)
 Siphon employs a set of algorithms for performing virtual sink discovery and
selection, congestion detection, and traffic redirection,
• VSs are used to remove data events from the sensor network when any symptom
of traffic load occurs.
• VSs act as siphons to tunnel out traffic from regions experiencing traffic overload,
as shown in Figure

6. Siphon:

7. Learning automata-based congestion avoidance in sensor networks (LACAS):
 A congestion avoidance scheme for sensor networks based on the concept of
learning automata (LA)
 Assumes that an automaton, which is a simple autonomous machine (code) capable
of making decisions, is equipped at every node in the network, as shown in Figure

 Only the intermediate nodes during a transmission have their automata working for
controlling congestion locally.
 At any time instant, if we observe the network topology, the automata stationed in
the intermediate nodes, and not the ones in the source nodes, act as congestion
controllers of data arriving from source nodes.
 Also, each of the nodes is independent in the network for controlling congestion.
 For the input to the automaton at time t = 0, the number of actions associated with
an automaton is limited to ﬁve, based on the rate at which an intermediate sensor
node receives the packets from a source node.
• The actions are denoted as ψ ={ψ1, ψ2, ψ3, ψ4, ψ5}, as shown in Figure
• The rates, ψ, that are inputs to an automaton stationed in a particular node, are based on the
number of packets dropped till then in the concerned node.

 At any instant, the choice of an action by the automaton, is rewarded or penalized by
the environment.
 At t = 0, all of the actions have equal probability of being selected Pψi
 Assuming the automaton selects ψ1 initially, based on the probability values of all the
actions at time t = 0
 The chosen action then interacts with the environment which examines the action ψ1
and rewards/penalizes it based on the number of packets dropped at that node.
 If the action ψ1 is rewarded, the probability of ψ1 is increased and the probability of
the other actions, i.e., ψ ={ψ1, ψ2, ψ3, ψ4, ψ5}, are decreased as:

 Alternatively, if ψ1 is penalized, the probability corresponding to ψ as well as for the
rest of the actions, {ψ2, ψ3, ψ4, ψ5}, will remain unchanged
 At every time instant, the system must satisfy
 Probabilities of all the actions are updated continuously until the most optimal action
is selected, i.e. implying the probability corresponding to the most
desirable action tends to unity, as time approaches inﬁnity
 Assuming t → ∞, the automaton selects action ψ2 as the most optimal action for the
system, the sensor nodes emit the packets with rate associated with the action ψ2

8. Ant-based routing with congestion control (ARCC):
 Uses the concepts of ant colony optimization (ACO) to deal with congestion in a WSN
 Finds an optimum path between a source and sink by observing the network
performance issues such as throughput, fairness, and loss of packets
 Motivations behind ARCC:
• Nodes are assigned different levels of priorities according to their role and location. Thus, the
congestion control mechanism has to assign weighted fairness nodes, according to priorities.
• As the number of communicating nodes vary with time, a previously determined route may not be
best always. With congestion control, single path can be declared as most efﬁcient offering
minimum trafﬁc times during peak hours.
• ACO routing protocols require to consider QoS metrics to enhance overall network performance
• Maintaining routing tables is an overhead for sensor nodes. Therefore, ARCC mitigates the
overhead by running the ARCC algorithm every time, eliminating the need tp keeping past records.

8. Ant-based routing with congestion control (ARCC):
 To deal with node mobility, ARCC uses ACO and congestion control every time a
node communicates with the sink or any of its parent nodes.
 Assumes that all the sensor nodes cannot take actively participate in the data
communication process.
 Nodes with diminishing power sources or with other constraints may behave
selﬁshly at any time.

Reliability-centric protocols
• Packet reliability implies successful delivery of a packet from source to sink.
• In WSN, reliability may mean packet reliability, event reliability, end-to-end
reliability, hop-by-hop reliability, upstream or downstream reliability
• Event reliability is achieved when a detected event is reported to the sink with a
certain degree of accuracy
• End-to-end reliability refers to the successful data delivery from source to
destination
• Hop-by-hop reliability means reliable data delivery from a source to its next hop
• Upstream or sensors-to-sink reliability means reliable delivery of data from sensor
nodes to the sink,
• Downstream or sink-to-sensors reliable delivery ensures delivery of data or query
from the sink to all (or a subset of) sensor nodes.

1. Event-to-sink reliable transport (ESRT)
• For reliable detection, a sink relies on aggregated data provided by multiple source
nodes, not a single node => conventional end-to-end reliability not required for WSN
• A transport protocol for reliable event detection using minimum energy. Congestion
control mechanism is also added for achieving reliability and saving energy
• For reliable temporal tracking of an event, the sink evaluates the event features every t
time units where t represents the decision interval.
• Number of received data packets is used to evaluate reliability of event features
transportation from source nodes to sink

2. Reliable multi-segment transport (RMST)
• RMST is a selective NACK-based transport layer protocol designed to support directed
diffusion ensuring guaranteed delivery and, if required, fragmentation and reassembly
• Reliability mechanisms can be implemented in the MAC layer, transport layer,
application layer, and/or any combination of these layers.
• Main responsibility of RMST is delivery of any or all fragments of a unique RMST entity
to all concerned sinks.
• A unique RMST-entity is a data set that may or may not be fragmented into multiple
pieces, originating from the same source.
• Two distinctive transport services are provided: guaranteed delivery, and effective
fragmentation and reassembling of messages.
• RMST operates in two modes: caching and non-caching, which are conﬁgurable at the
run time.

3. Reliable bursty convergecast (RBC)
• In event-driven communications, there is a sudden increment of packets, generated
and ﬂown through the network, within a short span of time.
• Such high volume data amplify the channel contention and, as a consequence, the
packet collision probability is also enhanced.
• In a multi-hop network, the probability of packet collision is increased further
• The RBC protocol applies window-less block acknowledgement and differentiated
contention control mechanisms to overcome the challenges of reliable and real-time
bursty convergecast.
• The window-less block acknowledgement scheme helps RBC to forward packets
continuously in the presence of packet and acknowledgement loss,
• Alternatively, the differentiated contention control mechanism reduces the channel
contention by ranking the nodes on the basis of their queues and en-queued packets.

4. Pump slowly, fetch quickly (PSFQ)
• In PSFQ, packets from the source nodes are pumped at a relatively slow rate.
• The nodes that experience packet loss are allowed to fetch the missing packets
relatively quickly from the immediate neighbors that have copies of it.
• The losses are detected when a node receives a message with a higher sequence
number than is expected.
• PSFQ has three components:
i. Message relaying (pump operation)
ii. Relay-initiated error recovery (fetch operation)
iii. Selective status reporting (report operation)

5. GARUDA
• GARUDA addresses the problem of reliable downstream, point-to-multipoint data
delivery, i.e. delivery of data from the sink node to the source nodes.
• GARUDA is reliable and is scalable with the increase in the size of the network,
message characteristics, loss rate, and reliability semantics
• The reliability semantics are defined according to the following classifications:
 Reliable delivery to all nodes in the field.
 Reliable delivery to a part of the field.
 Reliable delivery to minimal number of sensors that can cover the field.
 Reliable delivery to a probabilistic subset of the sensors in the field.

5. GARUDA
• Main components of GARUDA are:
• The core nodes are used to cache the packets and the noncore nodes are used to
recover the lost packets
a. Construction of loss-recovery servers (core)
b. Loss-recovery process.

6. Asymmetric and reliable transport (ART)
• Aims at providing event reliability, instead of per-message reliability.
• The protocol is designed on the careful observation that there exist a lot of redundant
message transmissions in WSNs
• Depending on type of data aggregation used at intermediate nodes in the multi-hop
path, copies of messages of the same event that travel towards the sink are reduced.
• ART considers two types of reliable data transmissions: event reliability and query
reliability
• ART further classiﬁes the nodes in a sensor ﬁeld into essential (E) nodes and
nonessential (N) nodes.
• End-to-end reliable data transmission, in both the upstream and the downstream
directions, is capacitated with the help of asymmetric acknowledgement (ACK) and
negative acknowledgement (NACK) between the essential nodes and the sink node

6. Asymmetric and reliable transport (ART)
• ART has the capabilities of congestion control:
 Classiﬁcation of nodes into E and N is utilized to mitigate congestion when it occurs
 Congestion is monitored by the E nodes by monitoring the duration of ACK arrivals
corresponding to event messages.
 If within a pre-conﬁgured timeout interval, no ACK is received by an E node, the data
arrival at the N nodes is slowed down (or even temporarily stopped) by sending out
congestion alarm messages.

7. Collaborative transport control protocol (CTCP)
• A transport protocol aimed at providing end-to-end reliability.
• A collaborative protocol in which the nodes in the network collaborate to detect and
then mitigate congestion.
• Capable of delivery of packets to the application layer in the base station, even in the
presence of failure and disruptions in the network.
• Explicitly takes reliability and energy-efﬁciency issues into account.
• As a congestion control protocol, it is capable of limiting the rate of forwarding of
packets at nodes.
• Two main functionalities of CTCP:
1. Hop-by-hop connection open and close
2. Controllably reliable delivery

Other protocols: Sensor TCP (STCP)
• A generic protocol to be used at the transport layer of WSNs
• Capable of supporting many simultaneous applications in the same
network
• Provides application-speciﬁc reliability and congestion detection and
avoidance.

Module 6
Localization and Tracking

1. Understand the fundamental concept of localization in WSNs.
2. Describe localization algorithms in WSNs
3. Discuss target tracking by deploying sensor nodes

References
Chapter 7
(ISBN: 978-0-521-19247-7)

Introduction
• Localization: Refers to determining the location of a device in the absence
of additional infrastructure, such as satellites.
• Traditionally, the localization problem has its origin in robotics, where it is
necessary to locate a robot in action
• Information provided by WSNs is highly correlated w. r. t. space and time
• In forest ﬁre monitoring application, the aim is to trace the precise
location of ﬁre in order to take appropriate measures

Introduction
• The localization problem for WSNs is very different from other networks
since sensor nodes are small low-powered devices
• In WSNs, the nodes determine their geographic positions by using
externally aided localization or self-localization techniques
• Most of the nodes are static, and generally nodes determine their
positions in the network initialization phase

Introduction
• Solutions: Trilateration and triangulation are among the popular methods
for localization where relative measurements from three or four reference
nodes are utilized to calculate a node’s own location.
• Few other solutions are:
• Bounding box method
• Multidimensional scaling method
• Hop-count-based approach

Introduction
• Target tracking is another well-studied problem area in wireless networks
• Tracking of a target requires knowledge about its location at different times
• Tracking has applications in the military where the location determination
of an enemy vehicle is necessary
• General characteristics of various tracking systems are target location
reporting and collaborative processing to remove redundancy.

Introduction
• Node and target localization are necessary for successful target tracking
• Target tracking schemes related to WSNs are required to maintain balance
between various network resources such as power consumption,
communication bandwidth, and protocol overhead
• Using WSNs for target tracking offers advantages such as increased
observation quality and tracking accuracy, and system robustness.

Localization
• In WSNs, it is possible for a node to find its position within a few meters of
accuracy, if it is equipped with global positioning system (GPS).
• In practice, the use of GPS in WSNs with thousands of nodes is not feasible:
 Reason 1: A GPS receiver is expensive and the cost of deployment of WSN increases
if GPS is built inside every node.
 Reason 2: Use of GPS with every node is not an energy-efﬁcient solution in WSNs,
as these networks are intrinsically energy constrained.
 Reason 3: GPS does not work in indoor environments, and environmental factors
such as large buildings affect GPS performance

Localization
• In order to understand the problem of localization, let us consider a sensor
network deployed in a region, with N randomly deployed nodes.
• Assume that the position of node i is denoted as Pi, where i=1,2,..., N.
• The coordinate of Pi is presented as Pi=(xi, yi, zi)
• Some of the nodes such as beacon nodes, anchor nodes, or landmarks
know their positions by using GPS or some other methods.
• The nodes that do not know their positions at the beginning are known as
unknown nodes.
• Let the communication range of all types of nodes be R.
• If a node can directly communicate with another node, then the distance
between these two nodes is less than or equal to r, and these two nodes
are the neighboring nodes

Localization
• The localization problem is abstracted as follows:
 A WSN is represented by a graph G
=(V, E), │V│ = N
 An edge exist between two nodes,
if they can communicate directly
 B indicates the set of beacon
nodes with (xb, yb) for all b ∈ B is
given, and B ⊆ V
 The positions (xu, yu) for all
unknown nodes u ∈ U are to be
determined

Localization
• Two issues related with localization: (1) how to deﬁne the coordinate
system? and (2) how to calculate the distance between two nodes?
• Unknown nodes calculate their location by referencing certain number of
anchors using various ranging and direction methodologies

Localization: Distance estimation techniques
1. Received signal strength indication (RSSI)
 Energy of a radio signal decreases proportional to the square of the distance traveled
by it when the signal propagates through a medium
 So, a receiver can estimate its distance from the source by knowing the signal strength
at the source end and the strength of the signal at the receiver end
 A relation between the received power and distance:
Pr= kd-α
Pr is the power of the received signal,
k is a constant that depends on the frequency and transmitted power,
d is the distance between the transmitter and the receiver
α is the attenuation exponent
 Note that the RSSI measurements contain noise up to several meters in a few areas
such as indoors

2. Radio hop count
 If two nodes can directly communicate with each other, then it can be observed that
the distance between them is less than or equal to R, where R is the maximum radio
range of a node
 A WSN can be represented as an unweighted graph, in which the sensors are
represented by vertices and there is an edge between two nodes if the Euclidean
distance between these two nodes is less than or equal to R
 Length of shortest path between nodes si and sj is hij where hij is the hope count
between nodes si and sj
 If the distance between si, sj is denoted by dij, then dij is less than or equal to hij× R and
in the ideal case, this distance is equal to hij × R.

Performance metrics
2. Radio hop count
 In general, the distance between any two nodes depends on their spatial distribution
 The expected distance, which is covered per communication hop, is denoted as dhop.
 If nlocal is the expected number of neighbors, then the value of dhop be calculated as:
 Then dij can be calculated as dij ≈ hij × dhop
 The distance between any two nodes is calculated as integral multiples of dhop which
justiﬁes an inaccuracy of almost 0.5R in every measurement
 Note that the environmental obstacles may inﬂuence the connectivity of the graph

2. Radio hop count
 Example of hop count:
• Hop count of AE =3 and hop count of AF =3.
• Distance AE is less than distance AF

Localization
3. Time difference of arrival (TDOA)
 In the TDOA approach, each node is equipped with a speaker and a microphone.
 In the transmitter initiated approach, the transmitter first sends a radio message.
 After sending the message, it delays for some fixed interval of time, t, and then
generates some fixed patterns of sound by its speaker.
 When a receiver listens to the radio message, it records the receiving time, t, and
waits for the sound signal.
 It records the receiving time of sound signal as tradio and waits for the sound signal
 It records the receiving time of sound signal as tsound
 The receiver then calculates the distance d from the sender as follows:
where sradio and ssound are speed of radio signal and sound signal, respectively

Localization

Localization
 In another approach, a transmitter sends signal to multiple receivers with known
locations.
 The TDOA of a pair of receivers i, j is given as:
where tij is the receiving time of the signal at receiver i,
c is the signal propagation speed,
|| . || is the Euclidean distance

Localization
 Example of time difference of
arrival (TDOA) where a transmitter
calculates its position by using
location information from four
receivers

Localization
4. Angle of arrival (AOA), digital compasses
 In AOA approach, nodes are capable of
sensing the angle of arrival of a received
signal.
 AOA sensing requires an array of antennas
or multiple ultrasound receivers to be
equipped with the nodes.
 Each node in the network measures all
angles based on its main axis.

Localization algorithms
• Localization algorithms can be categorized as centralized and distributive.
 In centralized algorithms, a single powerful node calculates the positions of the
unknown nodes.
 An unknown node sends its measured information, such as beacon position and
distance, to the central node, and the latter node sends back the estimated position
to the former node.
 The problem with the centralized approach is that lots of packets are exchanged
among the central and the unknown nodes and, as a result, scalability is a real issue.
 If the network size increases, then the power consumption due to communication
also increases.

Localization algorithms
• Localization algorithms can be categorized as centralized and distributive.
 In distributive localization schemes, an unknown node estimates its own location

Localization: Centralized algorithms
• MDS-MAP
 A WSN is represented as an unweighted graph, where nodes are represented by
vertices and there is an edge between two nodes, if the distance between them is
less than or equal to the communication range of radio.
 The connectivity graph is assumed to be a connected graph that is a path between
any pair of nodes.
 The algorithm first produces a relative map of the network and then the relative map
is transformed into absolute positions.
 The task of finding a relative map has two phases:
 In an absolute map, the geographic coordinates of each node are determined
• In the first phase, the neighbors of each node are discovered and a connectivity graph is created.
• In the second phase, this graph is mapped onto a two- or three-dimensional plane.

• MDS-MAP
 Assuming there are some nodes with known positions in the 3D space, the straight
line distance between any such pair of nodes is known.
 Multidimensional scaling (MDS) can be used to map the nodes on a 2D plane where
nodes are placed on the basis of 3D distances among them.
 MDS starts with proximity matrices derived from points of multidimensional spaces
and then determine the placement of points on a low-dimensional space (generally
2D or 3D, where the distances among the points maintain original similarities)

• MDS-MAP
 Steps in the MDS algorithm:
1. Ranging data from the network are gathered, and a sparse matrix R is calculated.
Here, rij is the range between nodes i and j, or zero if no range is collected.
2. Shortest paths between all pairs of nodes in the region of interest are determined.
The shortest path distances are used to calculate the distance matrix, D, for MDS.
3. Classical MDS is applied on the distance matrix D. The 2D (or 3D) relative map is
constructed by retaining ﬁrst two (or three) largest eigenvalues and eigenvectors.
4. If there exists a sufﬁcient number of anchor nodes, the relative map is transformed
into an absolute map on the basis of absolute coordinates of anchor nodes.

• MDS-MAP
Beacon density vs. granularity of localization regions

• Adaptive beacon placement
 In distance-based approaches, the density and placement of beacons affects the
quality of localization.
 Every node needs to hear a minimum number of beacon nodes and those beacons
should be noncollinear.
 A uniform and dense distribution of beacons is not efﬁcient, although it may appear
that distribution improves the quality of localization.
 The hardware of beacons is costly; so, a larger number of beacons means increased
overall cost of the WSN.
 A high density of beacons implies the probability of collisions due to the transmission
also increases. So, a limited number of beacons are required for reducing the number
of collisions, saving energy and, hence, prolonging the lifetime of the network.

 Beacons are placed at known position (XB, YB), and they periodically transmit with a
time period t.
 Clients listen for time period t >>T.
 If the number of messages received by a client from a particular beacon exceeds a
threshold value, then that client is connected with the beacon. A client then
estimates its position (XE, YE), as it is the centroid of all connected beacons
 If the actual position of clients is (XA, YA), then the localization error LE is:
 As the density of beacon nodes increases, the size of localization area becomes ﬁner,
and hence, the localization error decreases

 In another approach for localization, solution improves incrementally by either
adjusting the position of beacon nodes, or adding some new beacons.
 This improvement is based on existing localization at any time instant.
 GPS equipped mobile robots are used for localization and estimate proper position to
deploy the beacons
 Assuming the area is a square, where the length of a side is S, each robot’s range is s
and transmission range of beacons is R.
 Three beacon placement algorithms are – Random, Max, and Grid

 Random
Step 1: A point ( Xr, Y) is chosen randomly.
Step 2: A beacon is placed on that point.
 The algorithm is used to calculate localization error of other algorithms
 MAX
Step 1: The terrain is divided into s × s squares.
Step 2: Localization error at each point is calculated. The coordinate of each point:
(i*s, j*s), where 0 ≤ i, j ≤ S/s. Number of data points in the terrain are:
Step 3: New beacon is added to point (X, Y) having maximum localization errors

 Grid
 In Grid, a candidate point is estimated by calculating the cumulative localization error
over each grid for several overlapping grids in the terrain.
Step 1: The terrain is divided into s × s squares.
Step 2: Localization error at each point is calculated. The coordinate of each point is
(i*s, j*s) where 0 ≤ i, j ≤ S/s.
Step 3: The terrain is divided into NG partially overlapping grids as follows.
Step 3.1: Each grid has a side, gridSide = 2R
Step 3.2 The center of grid G(i; j) is
GC(i, j) = (XC(i, j), YC(i, j))

 Grid
Step 4: For each grid G(i; j), the cumulative localization error LE(i; j) is calculated for
points measured in Step 2 that lie in the grid G(i; j). The number of data points
per grid is
where is the number of data points in the terrain
Step 5: The new beacon is added at the center GC(i; j) of the grid G(i; j) with the
maximum cumulative localization error

Module 7
Localization and Tracking

Contents
1. Localization.
2. Target tracking

Localization: Distributive algorithms
• Localization algorithms should mainly fulfill three conditions i.e. they
should be self organizing, robust, and energy efficient
• Sensor nodes are distributed randomly at the time of network installation
when the sensors are placed in an uncontrolled manner
• So, sensors should self-organize themselves in such scenarios.
• Nodes without GPS may find their locations in three phases:
Phase 1. Finding the distance between anchor nodes and node itself.
Phase 2. Calculating the location from the calculated distances.
Phase 3. Refinement of the location using information of neighbors.

• Beacon-based distributed algorithms: Diffusion
 In diffusion technique, an unknown node first finds the positions of its neighboring
nodes. Then, the unknown node estimates its position as the centroid of its
neighbors
 In another approach, each node estimates its position as centroid of its neighbors,
beacons and the unknown nodes. Unknown nodes run this process until the result
converges. The steps of this algorithm are as follows:
Step 1: All unknown nodes initialize their position as (0, 0).
Step 2: An unknown node finds the positions of all its neighbors.
Step 3: An unknown node estimates its position as the average of all its neighbors’
position.
Step 4: Steps 2, 3 are repeated until the result converges.

• Beacon-based distributed algorithms: Bounding box
 In this approach, each node listens to its neighboring beacon nodes.
 Assuming the position of a beacon is (xb, yb), and the communication range is r, if an
unknown node hears beacon, then it is located within a box, whose two corners are
((xb− r), (yb− r)) and ((xb+ r), (yb+ r)). This can be expressed as:
 The position of the node is within the intersection of all the bounding boxes
corresponding to all the neighboring beacons as:
where i =1,2,3... n, and n is the number of neighboring beacons

• Beacon-based distributed algorithms: Other Algorithms
 APIT a range-free area-based localization scheme
 Multilateration a distributive process of localization

Localization: Relaxation-based algorithms
• Anchor-free localization (AFL)
 AFL is a concurrent and anchor-free scheme to solve the localization problem
 Nodes try to estimate positions from local distance information where no node has
any location information using GPS or any other method.
 The estimated coordinate system is not unique and it can be mapped on a global
coordinate system in various ways by rotating, ﬂipping, or translating.
 Each node is assumed as a “point mass” and the nodes are connected with “strings”
 Force-direction relaxation methods were used in this localization scheme which
attains a minimum-energy conﬁguration of the nodes.

 The step-wise algorithm is as follows:
1. A node n1 is chosen at the periphery of the graph
2. A node n2 is selected such that n2 is maximum hop count away from n1
3. A third node n3 is selected such that n3 is maximum hop count away and equidistant
from both nodes n1 and n2
4. Node n4 is selected such that it is maximum hop count away from n3 and equidistant
from nodes n1 and n2
5. Node n5 is selected such that it is equidistant from each of nodes n1, n2 , n3 and n4
6. For each node ni, the hop counts h1i, h2i, h3i, h4i, h5i are estimated from chosen
reference points

 The algorithm is as follows:
7. For each node ni, the approximate polar coordinate (pi, θi) is estimated by using the
hop counts and radio range R
8. A local optimization technique is performed on current estimated coordinate by the
nodes.

Localization: Coordinate system stitching-based algorithms
• Robust distributed network localization with noisy range measurements
 In this approach, the nodes ﬁrst estimate the distances of all one-hop neighbors
which is exchanged among the neighbors.
 Using this information, each node localizes itself and its neighbors and then nodes are
organized into clusters where a cluster refers to a node and its one-hop neighbors.
 During this process, the nodes form a local coordinate system as they do not have
any knowledge about the global one.
 To transform the local coordinate system into global, the overlapped clusters are
merged or stitched.
 All sets of four nodes that are fully connected are found. These quadrilaterals are
taken as the smallest sub-graph and called “robust quad.”
 Relative positions of the nodes of a robust quad are unambiguous even in the
presence of measurement noise.
 Two robust quads are “chained” if they have three common nodes

 The algorithm has three phases, which is described as follows
 Phase 1 Cluster localization: Each node becomes the center of a cluster and finds all
its neighbors and estimates the relative location of all the neighbors. All robust quads
within a cluster are identified and the largest subgraphs with overlapped quads are
also identified. The position of a node within a cluster is then computed using
chaining of robust quads and trilateration.
 Phase 2 Cluster optimization (optional): In this phase, the estimated positions are
refined using numerical optimization techniques such Newton–Raphson. Any error
that accumulates in the computation is reduced. One of the advantages of this phase
is that no additional overhead is added to the protocol.
 Phase 3 Cluster transformation: Transformations among the local coordinate systems
of neighboring clusters are computed by selecting the set of nodes in common
between two clusters.

 Robust quadrilateral and its decomposition in four triangles

Localization: Hybrid localization algorithms
• Localization with limited number of anchors and clustered placement
 A distributed localization scheme composed of two different techniques namely
multidimensional scaling (MDS) and proximity-distance map (PDM)
 The advantage of this scheme is its reduced complexity over MDS
 The steps of the scheme are as follows:
 Step 1: Secondary anchors are selected in this step: kp is the number of primary
anchors and ks is the number of secondary anchors for each of the primary anchor.
Primary anchors send invitation containing its unique ID, counter, and the number ks.
Initially, the counter is set to zero. An ordinary node performs Bernoulli trial with
success rate of p upon receiving this message. If the outcome is true, then it becomes
a secondary anchor. Thus, the total number of anchors in the network is p = kp (1+ks)
 Step 2: The primary anchors send packets containing the coordinate and proximity of
the packet, i.e. the hop count of the packet. Secondary anchors do the same, except
that the coordinate value in the packet is left blank.

Localization: Hybrid localization algorithms
• Localization with limited number of anchors and clustered placement
 A distributed localization scheme composed of two different techniques namely
multidimensional scaling (MDS) and proximity-distance map (PDM)
 The steps of the scheme are as follows:
 Step 3: All of the nodes receive a packet containing the proximity value. If a node
receives more packets, then it stores it only for lower proximity value.
 Step 4: The proximity value is exchanged between the anchor nodes. After knowing
the proximity value for any pair of anchor nodes, secondary anchors localize
themselves using MDS.
 Step 5: Proximity distance mapping T is calculated using the proximity matrix P and
geographic distance matrix L: T =LPT (PP)T
 Step 6: Ordinary sensor nodes calculate their positions based on the stored proximity
vector ps and the position information of the anchors.

Localization: Other algorithms
• Radio interferometric positioning system (RIPS): Based on the concept of
interference between pairs of two senders and two receivers
• Error propagation aware localization: An error propagation aware
algorithm for precise cooperative indoor localization

Target tracking
• Tracking a target with the help of deployed sensor nodes creates a few
possible applications in both the civilian and military domains
• Advantages of applying target tracking with WSNs include improved
qualitative measurement, accurate and timely signal processing, and
increased robustness
• But there exist challenges for using WSNs such as limited battery power,
low bandwidth, short communication range of nodes, and limited
processing capability.
• Target tracking approaches mainly focus on ﬁnding a balance between the
energy consumption of the sensor nodes and tracking accuracy
• Problem with the centralized target tracking approaches is that they are
vulnerable to a single point of failure and not scalable

Target tracking
• Distributed approaches, on the other hand; increase computation and
communication cost in the networks
• Typically tracking a target is based on three steps – (1) node localization,
(2) target localization, and (3) target location update.
• Based on the number of targets to track, the existing approaches may be
divided in two categories – (1) single target tracking and (2) multi-target
tracking
• Figure shows a single target tracking scenario in which sensor nodes
alongside target trajectory are activated and the remaining nodes remain
in the sleep state

Target tracking
• Figure shows a single target tracking scenario in which the sensor nodes
alongside target trajectory are activated and the remaining nodes remain
in the sleep state

Target tracking: Single target tracking
• The proposed solutions are classiﬁed into ﬁve different approaches:
1. Tree-based tracking,
2. Cluster-based tracking,
3. Prediction-based tracking,
4. Mobicast message-based tracking
5. Hybrid tracking methods.

• Tree-based tracking
 This scheme introduces a concept called dynamic convoy tree based collaboration
(DCTC) for detection and tracking of mobile targets.
 The convoy tree is formed with the sensor nodes around the mobile target, and it is
dynamically maintained by adding or removing nodes as the target moves.
 The initial convey tree is formed when a target is ﬁrst detected in which a root is
selected that collects more information from sensor nodes to reﬁne the information
 As the target moves, some nodes in the tree are no longer needed and so they are
removed from the convoy tree.
 The root predicts the future movement direction of the target and the nodes in that
area are activated

Convoy tree reconﬁguration in DCTC with target movement

 The root of the convoy tree also needs to be changed as the movement of the target
progresses which helps optimizing the communication overhead between the nodes.
 Convoy tree reconfiguration is formulated as an optimization problem and optimal
solution is based on dynamic programming (o-DCTC) with maximum tree coverage
and minimum cost.
 There are two methods proposed for expansion and pruning of the tree, namely the
conservative scheme and the predictive scheme.
 Two tree reconfiguration schemes are sequential reconfiguration and localized
reconfiguration.

• Cluster-based tracking: Continuous object detection and tracking
 Clusters are formed to support collaborative data processing required by the sensor
nodes.
 There are two types of clustering approaches – static or dynamic
 CODA uses hybrid clustering for continuous tracking with low message overhead
 Assumes a network where nodes are divided into static clusters; one cluster head
(CH) is present in each cluster.
 The CH decides the boundary nodes of the cluster, by solving the convex-hull
problem using the Graham scan algorithm after receiving the location information of
all the nodes.
 These nodes are notiﬁed using messages from the CH.
 Based on the classiﬁcation, the boundary sensors are named Static-cluster boundary-
sensors (SBs) and the remaining nodes are named Static-cluster inner-sensors (SIs).

• Cluster-based tracking: Continuous object detection and tracking
 In next phase, static clustering scheme is utilized for boundary tracking of the object
in which the sensors send control messages to notify the CHs about the detection of
an object.
 Based on the number of clusters that detects the object, there may be different
cases by which the boundary of the object with each static cluster may be identiﬁed.
 The boundary sensors are then organized in a dynamic cluster (dynamic in the sense
that the number of nodes in a cluster is changed as the target moves)
 CHs fuse the boundary data and send the data to the sink node.
 The sink node collects the information from the CHs and can determine the whole
boundary of the object.

• Cluster-based tracking: Localized policy-based target tracking
 This scheme maintains a balance between the energy efﬁciency of the nodes and the
target tracking accuracy.
 Prudent use of sleep and wake-up mechanisms, network lifetime can be increased.
 The movement of a target is modeled based on the Gauss Markov mobility model.
 On detecting a target, the cluster head that detects it activates an optimal number of
nodes within its cluster, so that these nodes start sensing the target.
 A Markov decision process (MDP)-based framework is designed to adaptively
determine the optimal policy for selecting the nodes localized with each cluster.
 As the distance between the node and the target decreases, the received signal
strength (RSS) increases, thereby increasing the precision of the readings of sensing
the target at each node.

• Prediction-based tracking: Prediction-based energy saving scheme
 PES minimizes the number of nodes involved in object tracking, while putting the
other nodes into the sleep mode to save energy
 The problem of object tracking involves S number of sensor nodes tracking O number
of moving objects.
 The sampling time required is X seconds and the event update rate is 1/T.
 The goal is to minimize the overall energy consumption while maintaining an
acceptable missing rate (missing rate denotes the ratio of sensor nodes that fail to
report the detection to the total number of sensor nodes)
 In the PES scheme, the number of active nodes and the sampling frequency are
minimized, to optimize the energy consumption.

• Prediction-based tracking: Prediction-based energy saving scheme
 PES has three parts – (i) prediction model, (ii) wake-up mechanism, and (iii) recovery
mechanism.
 Using the prediction model, PES predicts the future movement of the target and
activates only those nodes.
 The sensor nodes are selected to be activated based on the energy and performance
in the wake-up mechanism.
 The recovery mechanism is used when the target is lost.

• Mobicast message-based tracking: HVE-mobicast
 Hierarchical-variant-egg-based (HVE) mobicast is a mobicast routing protocol
proposed for sensor networks with the goal of maintaining power efﬁciency
 A variant of multicast which decides the forwarding zone of a message.
 Overall method is divided in two phases – egg estimation and distributed HVE-
mobicast.
Phase 1: All sensor nodes estimate the variant-egg
Phase 2: A distributed algorithm is designed to adjust the size and shape of the
variant egg.

• Hybrid tracking method: Distributed predictive tracking (DPT)
 A hybrid of two different schemes: the cluster-based approach is utilized for
scalability, and the prediction-based approach offers a distributed and energy-
efficient solution.
 DPT also provides robustness against node failure.
 In DPT, the sensor nodes are randomly distributed over the area. These sensor nodes
are of the same type and the CH knows their IDs, location and energy level.
 To enhance energy efficiency, the sensors remain in sleep mode until they are
instructed by the CH to perform a sensing task.
 The target is first detected by the boundary sensors of a cluster.
 A target descriptor (TD) is used to maintain the information of the target which
contains target id, present location, next predicted location, and timestamp.

Target tracking: Multi-target tracking
• Hierarchical Markov decision process (HMDP) for target tracking (HMTT)
 Uses an energy saving scheme for sensor nodes based on a realistic mobility model.
 The target tracking framework is cluster-based, and uses a two-level Markov decision
process (MDP) to predict the target trajectories.
 Energy efﬁciency of the sensors is maintained by determining the optimal sleep time
of the sensors.
 The sensors are assumed to be deployed randomly over a two-dimensional ﬁeld and
they are divided into few clusters.
 A cluster head has three states – sensing, listening, and tracking. Target mobility is
driven by the Gauss Markov (GM) mobility model.

Module 8
Topology Management and Control

Contents
1. Topology management
2. Taxonomy of topology management
3. Topology control
4. Modeling WSNs
5. Simulation models
6. Modeling the behaviour of sensors and sensor networks

Contents
7. Simulation tools for wireless sensor networks (WSNs)
8. Performance metrics
9. Fundamental models

1. Discuss the notion of topology management for WSNs.
2. Describe the concept of topology control.
3. Model and simulate the behavior of sensors and sensor
networks.

Required Reading
1. Chapters 8 and 9 Principles of Wireless Sensor Networks,
M. Obaidat and S. Misra, Cambridge University Press,
2014. (ISBN: 978-0-521-19247-7)
Recommended Reading
1. Wireless Sensor Networks, Ian F. Akyildiz and Mehmet Can Vuran, John
Wiley & Sons, 2010. (ISBN: 978-0-470-03601-3)

Topology management
• Deriving a simple graph of node connected with inter-nodal links and
virtual relationships.
• Flat topology
• Nodes are handled equally. Such a topology is also called unstructured.
• However, it leads to very poor and uncertain network connectivity.
• Hierarchical topology
• Nodes are classified into groups or clusters, thereby forming a hierarchical topology.
• Sovereign approach of node organization, in which every cluster is represented and
managed by a cluster head.

• Taxonomy of topology management

Taxonomy of topology management
• The taxonomy of topology
management algorithms in WSNs
is
(a) Topology discovery
(b) Sleep cycle management
(c) Clustering.
Each of these categories has its own
set of algorithms, as shown in Figure
1.
Figure 1: Taxonomy of topology management algorithms .

• Topology discovery
• Retrieving the topological details from
the nodes of the network.
• A base station enquires about the
topological trivialities by broadcasting
packets to the network.
• Consequently, the nodes reciprocate
by sending packets to the base station
itself.
• TopDisc algorithm
• Accumulate the entire network
topology from the perspective of a
single node.
• The monitoring node sends the
“topology discovery request” packet
to all the active nodes See figure 2.
• direct response (B>A), (C>B>A),
(D>B>A),
• aggregated response (C>B), (D>B),
(B>A).
Figure 2: Example of topology.

• Colouring algorithms for finding the responding set (node labelling
method)
• TopDisc uses a colouring scheme to propagate requests to nodes and find the
responding set.
• The 3-coloring scheme- significance of each of the colours is as follows.
• White: These are the nodes that are yet to be reached.
• Black: The cluster heads are denoted by a black colour.
• Gray: These grey-coloured nodes are the one-hop neighbours of the black-coloured
nodes.
• The 4-coloring scheme
• It introduces a fourth colour, dark grey.
• To reduce the overlap between clusters.

• TopDisc responding mechanism
• Every node maintains its neighbourhood and other associated information,
which is as follows.
• A grey node stores information about its neighbouring black node.
• Every node is aware of the parent black node, i.e., the sender of the topology discovery
request.
• After each transmission, a black node waits for responses from its children.
• These are aggregated and transmitted to the immediate parent node.
• Thus, after a series of packet transmissions, the initiator node is
knowledgeable about the complete topology.

• Sensor topology retrieval at multiple resolutions (STREAM)
• The algorithmic aspects of STREAM.
i. The monitoring node broadcasts the topology discovery packet containing two
specialized parameters – virtual range and resolution factor.
ii. The monitoring node broadcasts a packet and turns black. It gets added to a set called
the minimal virtual independent dominating set (MVIDS).
a) Any node within a black node’s virtual range is red colour.
b) Nodes within the communication range are blue nodes.
c) White nodes are the undiscovered nodes.
iii. Black/red nodes discard packets that come to them.
iv. This process of packet dissemination continues till all nodes are black or red.
v. All black nodes get added to MVIDS. These nodes are responsible for aggregation of
information and subsequent transmission of the aggregated information from their
children nodes.

• Sleep cycle management
• To manage and set optimal schedules of sleep and wake-up operations.
• Span: Span is a distributed, randomized sleep cycle management algorithm
that has its applicability in a dense wireless network.
• Span aims at the following.
• Each point in the network is covered by at least a single coordinator node.
• Coordinators are scheduled in a rotating fashion.
• It aims at selection of an optimal number of coordinators.
• Election of coordinators is locally managed.

• The key functions of the SPAN algorithm (refer to Figure 3) are as
follows.
i. Nodes maintain state information and proactively broadcast HELLO
messages.
ii. A node turns on its radio after a fixed interval .
iii. A coordinator node, backs out if two of its neighbours can communicate
without intervention.
iv. A grace period is the interval of time between withdrawal of one
coordinator and replacement of the other. Each coordinator node must
serve this period before going to sleep.
v. Span rotates and distributes the role of coordinators. This leads to
distribution of responsibility and reduction of energy exhaustion.

• Sleep cycle management algorithms
• Span
Figure 3: Span forwarding backbone formed by the coordinators (black nodes).

• Geographic adaptive fidelity
(GAF)
• An energy-efficient algorithm
incorporating location awareness by
using the global positioning system
(GPS).
• It can be analysed by the following key
features.
i. The entire network is viewed as
several square grids, see Figure 4.
ii. The master node is in charge of
managing the grid and reporting
data.
iii. One of the slave nodes volunteers
to be the master node. The master
node, however, does not perform
any aggregation.
iv. The possible set of states for each
node is discovery, sleeping and
active.
Figure 4: GAF virtual grids with.

• Thus, GAF maintains network connectivity
without degrading the routing fidelity. The
transition diagram of GAF is shown in Figure 5.
Advantages
• Routing is done in a distributed
manner.
• The idea of grid avails the advantages
of modularity.
• Energy management is done
intelligently.
Disadvantages
• Although it is energy aware, the use of
GPS decelerates the performance.
• The state transitions involve energy
expense.
Figure 5: GAF state transitions.

• Cluster-based energy conservation (CEC)
• CEC maintains three types of nodes, as shown in Figure 6.
a) Cluster head (CH): The usual notion of a cluster head (CH) exists.
b) Gateway nodes: These nodes connect clusters and act as cluster gateways.
c) Redundant nodes: Nodes in sleep state.
Figure 6: CEC cluster formation.

• Sparse topology and energy management (STEM)
• An alternative solution to idle listening.
• STEM is a two-state algorithm composed of the following.
a) Monitor state:
b) Transfer state:
STEM considers two kinds of nodes.
The initiator node
Target node.
• STEM channels
a) Wake-up
b) Data

• STEM has two implementation
versions.
• STEM-B: In STEM-B, a sender node
sends a beacon containing the
source and target address.
• STEM-T: In STEM-T mode, the
sender transmits a continuous
interrupt signal to wake up the
target node.
• In both STEM versions, nodes other
than the initiator and target are
kept in the sleep mode see Figure 7.
Figure 7: State transitions in STEM, f1: wake-up radio, f2:
primary radio frequencies

• Naps
• This protocol aims to find a subset of nodes that may turn their radio off for some
period of time.
• Naps deals with two kinds of nodes.
• Waking
• Napping
• The steps of NAPS are as follows.
i. The initiator node broadcasts a HELLO message and starts a timer.
ii. For each HELLO message that the node receives, it increments a counter, that was initially
set to zero.
iii. Step (ii) repeats until the timer times out or the counter hits a threshold.
iv. If the counter reaches the threshold, before the timer times out, the node naps till the
timer stops.

• Clustering
• Clustering algorithms introduce hierarchy into the network.
• Nodes are classified into clusters governed by a cluster head.
• Data packets from each member node of a cluster are transmitted to the cluster
head.
• The cluster head is responsible for aggregating the individual node data to a
composite value.
• Clustering algorithm Based on cluster formation strategy
• Static
• Dynamic
• Clustering algorithm Based on nature of the network resources
• Homogeneous clustering
• Heterogeneous clustering

• Homogeneous clustering algorithms
• A homogeneous WSN consists of identical resources.
• Homogeneous categories.
1) Signal-based clustering algorithms.
2) Distance-based clustering algorithms.
3) Neighbour-based clustering algorithms.
• Signal-based algorithms:
• Low-energy adaptive clustering hierarchy (LEACH):
• It works on the principle of dividing nodes into clusters. governed by a cluster head (CH).
• Nodes inside a cluster communicate directly with the CH.
• The CH is responsible for data fusion and subsequent transmission to the base station (BS).

• LEACH Phases
• Setup phase-
• In this phase, the clusters are organized and a cluster head is determined.
• At the beginning of every round, each node probabilistically elects itself to be the cluster
head.
• Steady-state phase-
• The nodes communicate with the heads by sending single frames in their slots.
• Each node strives to attain energy efficiency by turning on its radio before the time of
transmission.
• Access-based energy-efficient cluster algorithm (ABEE)
• Parameters
• Network lifetime
• deployment
• Node correlation.

• Energy-efficient clustering scheme (EECS)
• Phases of EECS
• Cluster head election-
• This phase involves a competition of candidate nodes of a cluster to become the cluster head.
• Only a node that has an optimum distance and weight metric from other member nodes is
picked up as a cluster head.
• Cluster formation-
• In this phase, the cluster heads broadcast special messages called HEAD_AD_MSG. Nodes
obtain information about the communication range from these packets and decide which
cluster to join.
• Data are directly transmitted by the member nodes to the cluster heads.

• The clustering protocol (CP)
• The clustering protocol (CP) aims at arranging nodes into disjoint clusters.
• Each cluster can be viewed as a circle with the cluster head as the centre and a radius of
unit communication range.
• CP is defined as a covering problem of hexagonal packing.
• Neighbour-based algorithms:
• Topology and energy control algorithm (TECA)-
• To increase network connectivity and lifetime.
• TECA follows the usual clustering approach such as CP, EECS, and LEACH.
• In other words, TECA establishes a connected backbone topology.

In TECA, five nodes states are
defined.
a) Initial
b) Sleeping
c) Passive
d) Bridge
e) Cluster head
• State transition diagram of TECA is
shown in Figure 8.
• The important phases of TECA are
as follows.
i. Cluster head selection
ii. Bridge selection
iii. Sleeping timeout
Figure 8: TECA node state transitions

• Power-efficient gathering in
sensor information systems
(PEGASIS):
• To optimize the number of
transmissions and receptions.
• PEGASIS aims to form a chain of
nodes.
• Figure 9 shows the formation of
chain, starting with node 0.
• At every time instant, the closest
neighbour is added to the chain.
Figure 9: Greedy chain formation of nodes in
PEGASIS

• Heterogeneous clustering
algorithms:
• Low-energy localized clustering
(LLC):
• It follows a two-tier architecture, in
which the sensor nodes are present in
the lower layer.
• All cluster heads are used to compute
an asymptotic equilateral triangle
Cluster
• This phase follows two different types
of algorithms:
• NLP-based approach.
• VC-based approach
Figure 10: Cluster heads forming equilateral triangle

Heterogeneous clustering algorithms:
• Energy-efficient heterogeneous clustered scheme (EEHC):
• To find optimal cluster heads in a decentralized manner.
• It considers the spatial density of nodes.
• It also increases the network lifetime and performance.
• It classifies nodes as super, normal, and advanced, based on their health conditions.
• Based on a node’s considerable attributes, the CHs are elected.

Topology control
• Topology control strives to
maximize network lifetime and
optimize nodal interference
during communication.
• Topology control focuses on two
important network aspects
• Network coverage
• Network connectivity.
(See Figure 11.)
Figure 11: Taxonomy of existing topology control
schemes

Topology control
• Network coverage can be broadly classified into three types
Blanket coverage
• Coverage with the highest granularity)
Barrier coverage
• Coverage with medium granularity)
Sweep coverage
• Coverage with lowest granularity).

Topology control
• Network connectivity
• Defines the strength of inter-nodal connections.
• Connectivity can be studied under two domains.
• Temporal domain-
• Spatial domain
• Connectivity under spatial domain is mainly studied under two
categories –
• Homogeneous network
• Heterogeneous network

Modeling WSNs
• Radio propagation modeling (RPM)
• Radio transmission between the sender and receiver sensor nodes.
• Barriers in the communication route will worsen the RF signal propagating.

Modeling WSNs
• Basic transmission loss model
• Two-ray ground propagation
• Lognormal shadowing
• Density function
• Energy modelling

Simulation models
• WSN simulation models can be
categorized into four types:
• Environment model
• Sensor node model
• User node model
• Communication model
• Environment model
• Nodes are used to define the
physical environment.
• See Figure 12 Figure 12: A environment model of a WSN.

Simulation models
• Sensor node model
• The nodes in this model detail the conditions of nodes including communication,
mobility, and routing schemes.
• The architecture of these nodes can be illustrated by Figure 13.
Figure 13: A general architecture
of a sensor node model.

Simulation models
• User node model
• This model acts like an interface
between sink nodes and the user.
The data packets are used by the
user to analyze the targets.
• The architecture of this model is
given in Figure 14.
• These nodes get sensor reports and
transmit them to the application
layer.
Figure 14: A general architecture
of a user node model.

Simulation models
• Communication mode
• This model is categorized into three types:
• Environment sensor communication
• Sensor–sensor communication
• Sensor–user communication

• Modeling the behaviour of sensors and sensor
networks

Modeling the behaviour of sensors and
sensor networks
• Figure 15 illustrates the sensor
node structure.
• The key requirements for the
implementation and design process
of WSNs are as follows.
1) WSNs should be designed as self-
organizing.
2) More cooperative processing
should be performed.
3) There is a need for some good
security mechanisms.
4) The protocols or algorithms used
in a WSN should be energy aware.
Figure 15: Overall structure of a sensor node.

Modeling the behaviour of sensors and sensor networks
• Self-organization
• WSN can be made self-organizing.
• Cooperative algorithms
• These algorithms are mainly used to decrease network traffic affected by data
aggregation and pre-processing.
• Security mechanisms
• Wireless sensor network applications and the operating conditions of the environment
influence the selection of security schemes for the network.
• Energy-aware requirement
• Sensor nodes typically employ microcontroller hardware that offers many schemes such
as dynamic power management (DPM) for power saving.

• Simulation tools for wireless sensor networks
(WSNs)

Simulation tools for wireless sensor networks (WSNs)
• NS-2
• It is an open source network simulation tool. This is a popular network simulator,
which is built using OTcl and C++ object-oriented programming language.
• GloMoSim
• This simulator is programmed using PARSEC, which is a C-based simulation language,
for sequential and parallel execution of discrete event simulation models.
• J-Sim
• This tool is also based on two languages, like the NS-2 simulator: JAVA and Jacl
(which is JAVA version of Tcl). J-Sim simulator framework has a general packet-
switching network, which permits various Internet protocols.
• SENSE
• SENSE is component based, but it is built using C++. SENSE simulator supports the
energy model and only some protocols, which include NullMAC.

Simulation tools for wireless sensor networks (WSNs)
• Visual Sense
• It offers an accurate radio model based on the energy propagation model.
• Prowler and Jprowler
• These simulation tools provide very accurate radio models.
• Sidh
• It can simulate networks with thousands of nodes much faster.
• Optimized network engineering tool 10.5 (OPNET)
• It is meant to study the performance of all communication networks.
• TOSSIM
• This was intended to facilitate the development of sensor network
applications.

Performance metrics
• Energy efficiency
• The energy efficiency refers to the number of data packets that can be sent
successfully by using the unit of the energy.
• Lifetime of the WSN system
• This may be defined by using:
• the period of the time until the required quality-of-service (QoS) of the application
cannot be provided.
• the period of the time until some nodes consume all the energy/power
• the period of time until the WSN becomes separated.
• Reliability
• In the WSNs, event steadiness is employed as a measure to show how consistently the
sensed incident could be sent to the sink node.

Performance metrics
• Coverage
• It refers to the overall space that can be observed by using sensor node
devices.
• Connectivity
• This performance metric can be used to assess the degree of the
interconnectivity of the WSN
• QoS metrics
• There are some WSN applications that require a specific QoS metric, such as
constant bit rate/real-time applications.

Fundamental models
• Traffic model
Continuous delivery
Event-based delivery
Hybrid delivery
Query-based delivery.

Fundamental models
• Energy models
• Reduce the consumption of power by sensor communications.
• It can save energy naturally, for instance, by turning off the transceiver for a
period of time.
• To reduce the amount of the communications in the network.
• This approach requires functions such as data aggregation and data
compression

Main Reference
1. Chapter 8 and Chapter 9 (Principles of Wireless Sensor
Networks, M. Obaidat and S. Misra, Cambridge University
Press, 2014. (ISBN: 978-0-521-19247-7)

Module 9
Wireless Mobile Sensor Networks

Contents
1. Need and use for mobile sensor node.
2. Coverage and mobile sensors
3. Network lifetime improvement

1. Discuss the need for mobile sensor nodes.
2. Elaborate the WSN coverage and mobile sensors.
3. Describe techniques to improve and increase the network
lifetime.

Required Reading
1. Chapters 11 Principles of Wireless Sensor Networks, M.
(ISBN: 978-0-521-19247-7)
Recommended Reading
1. A Complete Guide to Wireless Sensor Networks: From Inception to
Current Trends by Ankur Dumka, Sandip K. Chaurasiya, Arindam Biswas,
and Hardwari Lal Mandoria, CRC Press, 2019. (ISBN: 978-1-1385-7828-
9)

• Need and use for mobile sensor node.

Need and use for mobile sensor node
• In WSN the nodes sense their
surroundings and communicate
with the sink(s).
• WSN Types
• Static-
• WSN is referred to SWSN (See Figure
1)
• Stationary
• WSN is referred to MWSN
Figure 1: An example of multi-hop communication in a
single sink SWSN.

Need and use for mobile sensor node
• Funnelling/bottleneck effect
• Distant or boundary sensor nodes in a multi-hop communication scenario
communicate with the sink
• Increase of data packets
• The effect of this congestion is packet dropping and/or retransmission of
packets
• Hotspot problem
• The one-hop neighbours of a sink transmit more data to it than the other
nodes.
• So, the battery power of a sink’s one-hop neighbours is depleted more rapidly
than the rest of the network.
• The hotspot problem generates two subproblems – (i) sink isolation and (ii)
network partitioning.

• Coverage and mobile sensors

Coverage and mobile sensors
• Applications of WSNs can be broadly categorized into two types:
• Monitoring
• Surveillance
• In the coverage problem, the number of sensor nodes needs to be
optimized.
• Networks Coverage Types
• Area coverage
• Target coverage or point coverage
• Barrier coverage

Coverage and mobile sensors
• Wireless sensor networks with mobile nodes can be classified into
two broad categories
• Hybrid WSNs
• Hybrid WSNs consist of both mobile and stationary nodes
• Mobile WSNs.
• Mobile WSNs are built from mobile nodes only.
• The strategies for mobile node deployment depend on the principles
of virtual force, computational geometry, and grid-based approaches.

Voronoi diagram-based approaches
• The Voronoi diagram is an important computational geometric
structure.
• It has several important applications in physics, astronomy, robotics,
and many more fields.
• It provides proximity information about a set of geometric nodes or
points.
• The Euclidean distance of two nodes p, q is denoted by ed(p, q).
• Let us consider a set of sensor nodes S deployed in a field. The
position of the ith sensor node is denoted by pi.
• The Voronoi diagram of S is the partitioning of the field into n cells;
each cell corresponds to one sensor.

• A point q is in the cell
corresponding to the ith sensor, if
ed(q, pi) < ed(q, pj), for all nodes ∈ S
and i ≠ j.
• In the left-hand part of Figure 2, a
Voronoi diagram for three points or
three sensor nodes is shown.
• An example of Voronoi polygon is
shown in the right-hand part of
Figure 2.
• The vertices of the polygon are v1,
v2, v3, v4, and v5.
• When sensor node s0 of Figure 2
tries to calculate the position of v1,
it requires the location information
of two neighbours, s1 and s5.
Figure 2 Example of Voronoi diagram (left) and Voronoi polygon
(right)

• Let us assume that the location of sensor node si is (xi, yi), and the line which
is perpendicular to the line connecting si, sj is Lij.
• The gradient of L05 is – ((x0 − x5)/(y0 − y5)) and the coordinate of the middle
point of the line connecting s0 and s5 is ((x0 + x5)/2, (y0 + y5)/2).
• From the above information, the equation of L05 can be estimated and the
equation is represented by equation (1).
• the equations of L05 and L01, s0 is able to estimate the position of point v1:

• A Voronoi diagram-based node-deployment protocol is illustrated to
optimize network coverage.
• Initially, all sensor nodes are deployed into the region of interest and
the sensors broadcast their locations.
• Each sensor node calculates its Voronoi polygon from the received
neighbourhood information.
• In the next step, the sensor nodes check whether a coverage hole
exists in their respective Voronoi polygon or not.

• The Voronoi-based algorithm (VOR)
• In VOR, a sensor node first estimates the existence of
a coverage hole within its Voronoi polygon.
• If the furthest Voronoi vertex is not covered by the
sensor node, then the node assumes the existence of
a coverage hole within its Voronoi polygon.
• When such a hole exists, it then moves toward the
furthest uncovered Voronoi vertex to cover that
vertex.
• A sensor node S and its Voronoi polygon are shown in
Figure 3.
• B is the furthest uncovered Voronoi vertex of sensor
node S. So, S moves towards B to cover B.
• Sensor node S estimates its new location along the
path SB and it moves to A to cover B. The Euclidean
distance between A and B is equal to the sensing
range of a sensor node.
Figure 3: Example of sensor movement
in VOR

• The maximum moving distance for a sensor node is the difference
between half of the communication range and the sensing range of a
node. A typical scenario is shown in Figure 4.
Figure 4: Inaccurate Voronoi polygon due to incomplete neighbourhood information.

• Minimax algorithm
• A node moves toward the furthest Voronoi vertex (Vfur), but the node
estimates its target location to minimize its distance to Vfur.
• In the Mimimax algorithm, the target position, known as the Minimax point
and denoted by pm, is chosen in a way that it reduces the variance of
distances to the Voronoi vertices.
• This way, the nodes avoid the changing of a close vertex into the furthest one.
The circumcircle of three vertices Va,Vb,Vcc is represented by C(Va,Vb,Vcc) the
algorithm to estimate the Minimax point of a given Voronoi polygon is given
below.

• Centroid and dual-centroid schemes
• The centroid and dual-centroid schemes are based on the centroid of a
polygon.
• In the centroid scheme, a sensor node, at the beginning of each round,
calculates its Voronoi polygon from its neighbours' location information.
• If there is no coverage hole within the polygon, the node skips the coverage
enhancement procedures for that round.
• If the node finds any coverage hole, then it estimates the centroid of the
polygon by using equations (2) and (3).

• Centroid and dual-centroid
schemes
• The location of centroid of a
polygon with n vertices is denoted
by (Cx, Cy) and area of the polygon is
shown by A:
• After calculating the centroid of the
Voronoi polygon and choosing the
centroid as its new location
• The node moves to the centroid, if
there is some improvement of local
coverage at that location, otherwise
it does not move.

• Dual-centroid schemes
• Voronoi polygon
• Voronoi neighbour polygon.
• The Voronoi polygons of sensor
nodes are shown in solid line and
Voronoi neighbour polygon of node
A is shown by dashed line in Figure
5.
• After calculating the centroid of the
Voronoi polygon and choosing the
centroid as its new location, the
sensor node estimates the
improvement of local coverage at
that new location.
Figure 5: Voronoi polygon and Voronoi neighbour polygon
of a node.

• Bidding protocols
• Bidding protocol has three phases
• Service advertisement-
• Each mobile node advertises its location and base price.
• Bidding advertisement-
• Static nodes detect coverage holes by scrutinizing their respective Voronoi
polygons.
• Serving advertisement-
• a mobile node selects the highest bid among the received bids and moves to
that area.

• Virtual force-based approaches
• Used to optimize the coverage area by the sensor network.
• Here nodes are assumed as virtual particles or electrostatic particles.
• Owing to the effect of repulsive force, nodes move away from one another
and the obstacles.
• This force is inversely proportional to the nodes’ distance.
• There is also an attractive force, called the viscous friction force, which helps
the nodes to reach a static equilibrium state.

• Virtual force-based approaches
• In this approaches nodes spread from a densely populated area to all over the
monitored area.
• Each node repels its neighbouring sensor nodes and is repelled by the local
obstacles.
• The network reaches a static equilibrium state when all nodes stop due to the
viscous force.
• It is assumed that, nodes are capable of finding nearby sensor nodes as well
as obstacles through communication or sensing.

• Grid-based approach
• Scan-based movement-assisted sensor deployment (SMART) is applied to unevenly
distributed sensor networks to balance the sensor distribution.
• Here, all the nodes are assumed to be mobile. The monitoring area is partitioned
into n × n grids or mesh of clusters.
• The number of nodes in each grid cell is assumed to be the load of that cell.
• Initially, sensor nodes are randomly distributed and the loads of the cells are not
equal.
• SMART helps to deploy the sensor nodes evenly and, hence, balances the load of the
cells. The basic principle of SMART comes from a two-dimensional scan-based
approach.

• Grid-based approach
• The balancing is performed by a two-round scan; the first one balances the
rows, and the second balances the columns.
• An example of a two-round scan is shown in Figure 6. In Figure 6b, the scan
process balances all the rows and in Figure 6c, all the columns are balanced,
and, hence, the total area is balanced.
Figure 6: Example of a two-round scan of SMART.

• Event coverage
• Mobile sensors, with collaboration of stationary sensors, are used for reliable
detection and location estimation of events.
• The mobile sensors perform two different tasks. Either the mobile sensors
monitor the area sporadically or, when static sensors inform the nearby
mobile sensors about a suspected event.
• The mobile nodes autonomously plan their paths on the basis of local
information such as their own measurements and information collected from
the neighbouring static and mobile nodes.
• The objectives of path planning are to reach the target area as fast as possible
and to improve the area coverage.

• Event coverage
• The sensing range of a static sensor
is divided into two parts
• The detection range
• sensor is able to detect the event
reliably and report to the sink about
that event.
• The suspicion range
• if an event occurs within the suspicion
range of a static sensor, it transmits its
suspicion, by a suspicion message,
either to the sink or to the nearby
mobile sensors.
• An example is shown in Figure 7.
Figure 7: Collaboration of stationary nodes and mobile nodes.

• Network lifetime improvement

Network lifetime improvement
• Predictable and controllable mobile sink
• The mechanism that considers the mobility of sink(s) and routing strategy of
sensor nodes, jointly, to increase the lifetime of a WSN.
• It is assumed that the sensor nodes are densely deployed by a Poisson
process within a circular area of radius r.
• The load of the ith node, loadi, represents the power consumed by node i
during transmission and reception of data. Higher load implies shorter
lifetime of a node.
• The network lifetime is, roughly, inversely proportional with the “network
load,” loadN.

• The load balancing problem is formulated as a min-max problem in
terms of the average load of sensor nodes:
• The average load, loadi, of the ith sensor node depends on the
routing strategies, R, taken by the nodes and the mobility strategies,
M, taken by the sink.
• The average load of a sensor node decreases with the increase in the
distance between the node and the sink.

• Finding the optimum joint mobility and routing strategies is a two-phase
process.
• In phase 1, the optimum mobility strategy of a sink is estimated by fixing the
routing of sensor nodes to the shortest path routing.
• After the estimation of the optimum mobility strategy, a better routing
strategy than the shortest path routing is sought.
• Only the periodic mobility strategies with finite period are considered at the
time of the estimation of optimum mobility strategy.
• The optimum symmetric trajectory is a circular trajectory around the centre
of the network. The load of the network is minimized when the radius of the
circular trajectory is equivalent to the radius of the network.

• To distribute the load on the nodes,
the network is partitioned into two
parts, as shown in Figure 9.
• The sink moves in a circular path of
radius rm around the centre of the
network, where rm < r.
• This path divides the network into
two parts:
• The circular area of radius rm
• The annulus between the boundary of
network and the circular path of sink.
Figure 9: Example of routing by nodes using joint
mobility and routing .

• A sensor node uses shortest path routing if it is within the circular area
enclosed by sink’s trajectory.
• When node S1 tries to communicate with base station B, it uses the shortest
path routing.
• The nodes in the annulus transmit packets using a two-step routing process
called “round routing.”
• A packet is transmitted along a circular path around the centre of the network
C until it reaches CB.
• Then the packet is forwarded using shortest path routing. Nodes S2 and S3
use round routing to communicate with B.
• The value of rm is determined from simulation to be roughly 0.9 × r.

• Predictable but uncontrollable mobile sink
• A framework for saving power of individual sensor nodes of WSNs in the
presence of a predictable, but uncontrollable, mobile sink or “observer.”
• The sensor nodes are distributed over the area of interest A.
• Two different kinds of distribution are considered here.
• In the first type, the sensor nodes are distributed randomly and uniformly
over A while, in the second, the minimum distance between any two sensor
nodes is d.

• Predictable but uncontrollable
mobile sink
• The sink S, with speed v, follows the
same path repeatedly. All the
sensor nodes are identical, and
their communication range is RC.
• Let us assume that each sensor
node requires tdata time for
transferring its sensed data to sink.
• A sensor node fails to communicate
successfully with a sink if the sink
does not stay within the node’s
communication range for at least
tdata time.
• An unsuccessful communication is
called an “outage”. See Figure 10.
Figure 10: A WSN and path of a predictable but
uncontrollable sink.

• Single-hop communication is used between a node and the sink. The sink
must come within the communication range of every sensor node during the
journey along its path within A.
• Let us assume that the maximum distance between any sensor node from the
path of the sink is D.
• So, for successful communication between the sink and any sensor node, the
communication range, RC, of sensor nodes can be calculated by using:
• The value of RC that satisfies equation (4) ensures that the sink will remain
within the communication range of every node for at least tdata time. The
relationship between D and RC is shown in Figure 11.

• The relationship between D and RC is shown in Figure 11.
Figure 11: Relationship between RC and D.

• The data collection by a sink is formulated as a queuing problem. As the sink
forwards along its path, new sensors come within the range of sink while
some sensors, previously able to communicate with sink, disappear from the
communication range of the sink.
• In Δt time, the sink moves vΔt distance. Sensor nodes that are within 2DvΔt
area appear within the range of the sink. While a new node appears within
the communication range of the sink, the node has to wait if the sink is busy
to communicate with others.
• There may be multiple nodes within the 2DvΔt area. While the sink
communicates with one, others have to wait for their turn. Each sensor, i, has
a maximum waiting time and this waiting time depends on its distance, dpath,
from the path travelled by the sink.

• Predictable but uncontrollable
mobile sink
• The maximum waiting time of node
i can be calculated as follows:
• The relationship between maximum
waiting time of a node and its
distance from sink’s path is shown
in Figure 12. Figure 12: Relation between maximum waiting time of a node and
its distance from the sink’s path.

• If the distance d between any two nodes is a minimum, then no outage is
guaranteed while d satisfies equation 6.
• The sensor network lifetime can be partitioned into three phases:
• Start-up- the sink and the sensor nodes exchange information to get to know one
another.
• Steady- The sink acquires location information of sensor nodes during the start-up
phase. Based on gathered information and its own location information, the sink initiates
communication by sending wake-up calls to the sensor nodes that it estimates to be
within its communication range.
• Failure-detection- The sink can detect node failures if some nodes do not respond to
multiple wake-up calls.

• Unpredictable and uncontrollable sink
• In, proposed a mechanism where stationary sensor nodes can communicate
with a randomly moveable sink.
• If the sink changes its location frequently and unpredictably, either sensor
nodes may communicate with the sink through flooding.
• Two protocols, local update-based routing protocol (LURP) and adaptive local
update-based routing protocol (ALURP), were proposed by Wang et al. [18].
• In the initial phase of node deployment of the LURP and ALURP protocols, the
sink broadcasts its location information within the whole network.
• Whenever a distant sensor node from the sink communicates with the latter,
communication is divided into two phases.
• In the first phase, the communicating node forwards the data towards a small
area known as the destination area, encircled around the sink

• Unpredictable and uncontrollable
sink
• An example of two-phase packet
forwarding mechanism of LURP is
shown in Figure 13.
• In ALURP, when a sink changes its
location within the destination area,
it does not broadcast its location
information to the whole
destination area.
• It restricts its update area and
creates an adaptive area. Figure 13: An example of message passing from a sensor node
to sink in LURP.

sink
• The radius of the adaptive area is
the distance between the virtual
centre (VC) and the sink’s current
location.
• An example of communication
between a sensor node and sink in
ALURP is shown in Figure 14.
• When a sensor node B, within
destination area A, receives a
packet to forward to the sink, it
forwards the packet to a node
• DNAA, which is inside the adaptive
area AA.
Figure 14: An example of message passing from a sensor node
to sink in ALURP.

• The node DNAA then forwards the packet to the sink.
• The sink’s initial location is its VC and the radius of adaptive area is zero.
• The size of the adaptive area increases as the sink moves away from the VC.
• The sink broadcasts its updated location information among all the nodes of
the adaptive area only and the nodes of the adaptive area update their
routing topology to the sink.
• Whenever a distant sensor node tries to send packets to the sink, the packets
first reach any node within the adaptive area.
• The node acts as a dissemination node and forwards the packets to the sink.
• A problem occurs when the sink moves towards the VC. The size of the
adaptive area shrinks. An example is shown in Figure 15.

sink
• As shown in Figure 15, the node
DNP1 is not in C, but it was in P. So,
when DNP1 receives a packet, it will
erroneously forward that packet to
DNP2, instead of DNC, as it has
obsolete location information about
the sink.
• To eliminate this problem, the sink
informs the nodes that are not in C,
but were in P, to remove the
obsolete routing information about
the sink.
Figure 15: The size of adaptive area is reduced due to the
sink’s movement towards VC .

• The sink collects data periodically from each sensor node. Each round of periodic
data collection has three phases.
• In the first phase, the sink broadcasts its location information among sensor nodes.
• In the second phase, all the sensor nodes send their sensed data using multi-hop
communication.
• In the third phase, the sink estimates its next location from the received residual energy
information of the network and reaches the estimated location before the next phase of data
gathering begins.
• The mobile sink moves according to the half-quadrant-based movement strategy
(HUMS). Each data packet contains three types of data;
• The first is the sensed data,
• The second is the residual energy and location of the sensor node with the highest residual
energy along the path from source sensor node to sink, and
• The third is the residual energy and location of the sensor node with lowest residual energy
along the above-mentioned path.

sink
• In HUMS, the sink generates a
coordinate system, taking its current
position as the origin of that system.
• The coordinate system is divided into
eight half-quadrants, as shown in
Figure 16.
• Case 1: distant MoveDest
• Case 1a: If there are no quasi-
hotspots present in the DestSector
• Case 1b: If the DestSector and at least
one of the forward sectors are clean
• Case 1c: If the DestSector is clean, but
quasi-hotspots are present in both
the forward sectors.
Figure 16: Different scenarios of half-quadrant-based moving
strategy.

• Case 1d: If DestSector is miry and at
least one of the forward sectors is
clean.
• Case 1e: if quasi-hotspots are present
in the DestSector and both the
forward sectors.
• Case 1f: If all the eight sectors have
quasi-hotspots
• MIPS (minimum-influence position
selection algorithm): The mobile
sink selects its destination point by
using MIPS within the sector
selected by HUMS Shown in figure
17. Figure 17: Influence of quasi-hotspots on a candidate
position.

• Case 2: adjacent MoveDest: If the
mobile sink is within the
communication range of MoveDest,
the sink selects a suitable position
near the MoveDest to force the
MoveDest to forward other’s data
and to consume more energy.
• An example of this scenario is
shown in Figure 18.
• The workload of MoveDest in Figure
18a is more than that of MoveDest
in Figure 18b. Figure 18: Example of different workloads of MoveDest
depending on location of the sink.

• Mobile relays and data mules
• Mobile nodes may be used as relay
nodes to increase the lifetime of a
WSN.
• Mobile relay nodes are used in the
scenarios where the sink and the
sensor nodes are stationary.
• The responsibilities include sensing
the surrounding environment,
processing the data, and
transmitting sensed or received
data to the sink (see Figure 19).
• Assume that a WSN is partitioned
into two components, component 1
and component 2.
• These two components are
connected with the sink through
sensor node 1, node 2, respectively.
Figure 19: One mobile relay node inherits the responsibilities of
multiple bottleneck sensor nodes in different time periods.

• As node 1 and node 2 exchange all the packets between the sink and
component 1 and component 2, respectively, they will drain their energy
more rapidly than the rest of the WSN.
• As a consequence, the WSN will be partitioned when either of nodes 1 or 2
dies.
• A mobile relay node may reduce the burdens of those bottleneck nodes.
• A framework for improving network lifetime using mobile relay node. N
sensor nodes are distributed by a Poisson point process in an area of radius R.
• The sink, S, is located at the centre of the monitored area. The transmission
range of all the sensor nodes is assumed to be equal to unity, and we also
assume that the nodes transmit their data to the sink at a fixed rate.

• The initial energy of a battery
powered sensor node is denoted by
E. It is assumed that the sink and
mobile relay have unlimited energy.
• The transmission range and sensing
range of the mobile relay are the
same as the sensor nodes’.
• The static nodes are partitioned on
the basis of their distance from the
sink. A node belongs to set Pi, if it is
able to reach the sink in i hops.
• An example is shown in Figure 20.
Figure 20: Partitions of sensor nodes in the circular network.

• A joint mobility and routing algorithm
is proposed to maximize the network
lifetime.
• It is assumed that the network is static
and densely populated.
• The starting location of the mobile
relay is same as the sink’s location. The
mobile relay traverses around the sink
until it reaches the periphery of Q2.
• The path of the mobile relay encircles
the sink with concentric rings with
increasing radii. After reaching the
periphery of Q2, it stays at each point
of the path and relays messages to the
sink.
• The messages are forwarded by using
the aggregation routing algorithm
(ARA). Shown in Figure 21.
Figure 21: Message forwarding with ARA (adopted from [21],
with minor modifications).

• The mobile nodes are used as data
collectors. These nodes, called data
mules, collect data from the stationary
sensor nodes of sparse sensor
networks, buffer the data, and transfer
the collected data to the sinks or
access points at appropriate time.
• The mobility model adopted for the
mules is the random walk model.
• The mules communicate with the
sensors in the short range. The three-
tier architecture saves the power of
sensor nodes, as all the sensor
communications are short range. (see
Figure 22).
Figure 22: Three-tier MULE architecture.

• An energy-efficient data
collection scheme is proposed in
[22]. In the proposed scheme, the
stationary sensor nodes have
three states:
• Sleep
• Discovery
• Data transfer
(See Figure 23).
Figure 23: Three states of a stationary sensor node.

Main Reference
(ISBN: 978-0-521-19247-7)

Module 10
Communication, Error Control,
Time Synchronization, Naming and Addressing
and Cross Layer Issues

Contents
1. Communication in Wireless Sensor Networks
2. Error and Control Issues
3. Time Synchronization Issues
4. Naming and Addressing Issues
5. Cross-Layer Issues

1. Describe the notion of cross-layer optimization in WSNs.
2. Discuss the Error and Control Issues for WSNs
3. Explain Naming, Addressing and Time Synchronization Issues for
WSNs.

References
Chapter 9
Energy-Efficient Wireless Sensor Networks, Edited by:
Vidushi Sharma and Anuradha Pughat, CRC Press, 2018
(ISBN: 13: 978-1-4987-8334-7)

Introduction
• Sensor networks differ from traditional wired and wireless networks in
terms of computation capabilities, energy, size, and memory.
• Due to this, the physical layer demands energy efficiency in modulation and
coding schemes.
• Some challenges in other operative mechanisms include time
synchronization and naming and addressing of nodes.
• All the operative protocols in these areas have to be energy efficient and
less complex.
• In this week, we discuss some of the above challenges and outline the
physical layer design concepts along with the other operative mechanisms
of various protocols.

Communication in Wireless Sensor Networks
• A frequency band is used for communication because of the inefficiency in
using a single frequency to communicate.
• WSNs use some of the license free bands called industrial, scientific,
medicine (ISM) bands for which permission from an authorized body is not
required and can be used with other frequency bands.
• Demand of using the radio frequency (RF) spectra is rapidly growing due to
increasing number of wireless and mobile communication applications.
• The industry has reached the limits of the current static spectrum
allocation that leads to open challenges of dynamic spectrum allocation
which provides sparsely used spectrums to the users

Energy Saving Methods in Communication
• Duty Cycling Approach
 During communication, some of the nodes remain idle while others are active.
 Even when the nodes are idle, they consume energy; so, the better way is to use the
radio in the best possible mode.
 The fraction of the active time of a node in one cycle is called its duty cycle.
 It is better to put the current inactive nodes in sleep mode, whereas active nodes
can switch off the radio when there is no network job

• Data-Driven Approach
 These approaches can be mainly classified as data reduction and energy-efficient
data acquisition.
 Data reduction is the process in which the larger entity of the collected data from
sensors is converted into smaller useful entity so that at a later stage the same data
can be retrieved without any loss.
 An important concept reducing the size of the data and thus minimizing the power
consumption.
 Data reduction approach also concentrates on preventing the nodes from
transmitting data to the sink that reduces the transmission load on the node as well
as the communication and processing overheads at the sink side

• Mobility-Based Approaches
 In mobility-based approaches, no restriction on connectivity is required.
 The communication between wireless sensor nodes needs a radio connection as a
physical layer in which energy is consumed when the radio sends or receives data.
 Since the main aim of WSNs is the optimal use of their cost and energy, the design
of the physical layer of a WSN becomes very important in this context.

• Aspects of the Physical Layer
 Modulation and demodulation of the data are associated with the physical layer.
 The transceiver has three modes: idle, sleep, and active - key to effective energy
management is to switch the radio off when the radio channel is idle.
 It was suggested that there are two factors responsible for energy loss in a wireless
transmission: (1) the loss due to the channel and (2) fixed energy cost to run the
transmission and reception circuitry.
 Both the above factors have a relation with the hop distance
 Increasing hop distance incurs channel loss and as the number of hops increases,
the cost increases linearly which implies there should be a balance between optimal
hop distance and the amount of energy consumed
 Chances of the transmission success depend upon the modulation used so use
efficient modulation techniques in the physical layer.

• Aspects of the Physical Layer
 In case M-ary modulation, the transmitted energy increases for a fixed bit error rate
(BER), whereas the number of transmissions decreases.
 Since the “on” time of a transmitter is very short for higher modulation, these
modulation schemes are preferred as energy-efficient schemes, although their cost
is high.

• Communication Protocols
 There are two main communication protocols in the domain of WSNs
1. 6LowPAN
2. ZigBee.

• Communication Protocols: 6LowPAN
 The 6LowPAN (released in 2007 by Internet Engineering Task Force [IETF]) is an
open standard communication protocol.
 Consumes less power, minimum data rate, and needs low-cost personal area
networks (PANs).
 Combination of two-Internet protocol (IPV6) and low-power PAN.
 It can also be used with the relationship of time variance among the nodes in WSNs.

• Communication Protocols: ZigBee
 A popular a low-cost, low-power, advanced communication protocol for small
devices used for low-rate wireless personal area networks (LR-WPANs).
 ZigBee is preferably used in body sensor networks (BSNs)
 ZigBee is also applied in a mesh network of routers to relay data from different
patients to the access point (AP).
• BSNs are a sensor or group of sensors attached to a patient and a coordinator for collecting raw
data.
• Data is sent, analyzed, and processed in control devices through the network
• ZigBee coordinator as controller works with interrupt to reduce usage consumption in the
network in gathering the raw data, e.g. healthcare monitoring.
 The AP is connected to the Internet to allow collaboration of doctors, medical centers, and other
data centers that gather patient records, so that decisions can be made.

• Communication Protocols: ZigBee
 Consists of two layers:
 The data rates between 10 and 250 kbps over a 10–75 m range is easily
communicable using ZigBee network devices.
• Application support layer
• Network/security layer.

• Energy-Efficient Modulation Techniques in Physical Layer
 An appropriate modulation scheme is required for effective communication systems
and thus, the survivability and lifetime of WSNs.
 Choice of the correct scheme of modulation depends on the network traffic and
reliable communication in a WSN.
 Various modulation schemes make the channel capable of sending maximum data
over the unsecure channel with high security.
 Since all the modulation schemes are not energy efficient, it is very important for
WSNs to use optimum modulation in terms of energy efficiency and minimum error

 In amplitude modulation (AM), when the incoming signal is a sequence of 0 and 1
value, the modulation process is called amplitude shift keying (ASK).
 In frequency modulation (FM), when the incoming signal is digital, the modulation
process is called FSK.
 In FM, when 0.5 modulation index is used, it is called minimum shift keying (MSK).
MSK can of detect coherent and noncoherent signals and amplify power efficiently
 Incoming digital signal with phase modulation (PM) refers to phase shift keying (PSK)
 Table 9.1 summarizes a quick view of different modulation schemes.

Features of Various Modulation Techniques

 In relevant literature, different researches have reported different observations:
 For better understanding, Table presents coding, performance parameters, and
overall system performance of different modulation techniques such as BPSK, QPSK,
16QAM, 64QAM, M-ary phase shift keying (MPSK), M-ary quadrature amplitude
modulation (MQAM), M-ary frequency shift keying (MFSK), and 8PSK
• One of the effort shows that adaptively chosen modulation and coding scheme can provide better
system performance
• In another research, it is suggested that the binary modulation scheme with an effective start-up
power dominant condition is more energy efficient
• Abouei et al. (2011) observed the concept of green modulation over Rayleigh flat-fading channels
to ensure energy efficiency in WSNs.

Comparison
of
Various
Modulation
Schemes

Error and Control Issues
• Because of low-power communication constraints, error-prone links occur
in WSN channels making error control a prime importance for WSNs.
• Length of network lifetime can be increased by putting the sensor node
radios to sleep as and when possible.
• For reliable data communication, two main error control strategies are:
 Transmission techniques must be chosen to utilize the active time of a sensor
node effectively.
 The design of energy and latency-efficient error control schemes play an
important role.
 Error-control ensures correct transmission and has a control on possible errors.
 Automatic repeat request (ARQ)
 Forward error correction (FEC)

Error and Control Issues: Aim of the Error Control
• The main aims of the error control is to ensure that data transport are:
1. Error-free and transmit exactly the sent bits
2. In-sequence and to send them in the original order
3. Duplicate-free and should be lossless

Error and Control Issues: Error Control Approaches
• Error control can generally be realized by backward error control (ARQ),
FEC, or a combination of the two, i.e., hybrid automatic repeat request
(HARQ)
• Automatic Repeat Request:
 In ARQ, the sender node adds error detection codes called parity bit to the data.
 Sink node checks the correctness of the received data.
 If there is an error in the received packet, the sink node rejects it and requests the
sender node to retransmit the same packet.
 The ARQ strategy results in latency and excessive energy cost

• Forward Error Correction:
• Also known as channel coding
• Error correcting codes are utilized to add redundancy to the packet for detecting bit
errors and corrects them at the receiver end.
• The transmit power required for BER or frame error rate can be minimized but this
leads to a high cost of extra energy consumption in encoding, decoding, and
transmitting redundant bits.
• Typically, energy spent on encoding is negligible while the decoding process
consumes significant energy implying FEC can be used in situations where
retransmissions are relatively costly

• Forward Error Correction: Error Correcting Codes
Block codes
 Block codes are of a fixed length nC, with nC–k parity bits, and are decoded one block
or codeword at a time, where k is the length of the information sequence.
 Hamming code (HC) is one of the basic codes introduced by Richard Hamming in 1950
 Over the years, more efficient and powerful codes are developed such as Reed–
Solomon (RS) and BCH. The RS, BCH, and HCs are the most widely known block codes.
 Balakrishnan et al. (2007) found BCH code outperforms over any other codes in terms
of energy efficiency requiring much less encoding/decoding energy consumption
 Goldsmith (2005) described a set of cyclic, linear, powerful BCH block codes for
moderate to high signal-to-noise ratio (SNR), generally outperforming all other block
codes at high rates.
 Short block codes like HCs can be decoded by syndrome decoding.

Convolutional codes
• Convolutional codes: For a rate k/nC, input is k bits, and output nC bits at each time
interval, but are decoded in a continuous stream of length L > nC.
• Encoding is performed in a continuous fashion rather than accumulating k data bits
and then encoding into n-bit code word as in block codes.
• A code word depends on both current k data bits and also on some earlier bits.
• The number of shifts a particular bit can influence output depends on constraint
length.
• Convolutional codes are decoded on a trellis using either Viterbi decoding, MAP
decoding, or sequential decoding.

Other codes
 In addition to traditional block codes and convolutional codes, there exist yet more
powerful codes such as turbo codes and low-density parity-check (LDPC) codes.
 All these codes have limited applications because of their computational complexity.
 Stronger codes are optimal to be used with end-to-end error control strategy while
simple codes are best for node-to-node error-control strategy.

• Hybrid Automatic Repeat Request:
 It is a challenge in WSNs to choose an optimum ECC keeping both the performance
and energy consumption in mind.
 ARQ provides reliable communication through retransmissions, which will be costly
in poor channels where retransmissions occur frequently.
 FEC performs better in poor channels, while the redundant bits become an undesired
cost when channel conditions are good.
 Some researchers have studied HARQ schemes, which include the advantages of
both error correcting schemes by combining ARQ and FEC.
 The FEC-based HARQ scheme with Bose–Chaudhuri–Hocquenghem (BCH) codes was
developed which is not good for all applications in WSNs, but it is limited to only
specific applications and consumes a large amount of energy.

Error and Control Issues: Challenges in Error Control
 Error control based on FEC or HARQ have the advantage of correcting a
certain number of errors in a packet. To avoid retransmissions, lower
packet error rate (PER) induced by FEC or HARQ could be used
 Lower PER could be managed by making longer hops in a multihop
network. By using FEC or HARQ, one extra hop can be avoided.
 By making control on transmission power, one can get desired output in
terms of energy efficiency as well as low cost.
 ECC is not always a practical and intelligent solution for increasing link
reliability. In some applications, an uncoded system may actually be
more energy efficient.
 Depending on application, analog decoders can be energy efficient in a
WSN. A combination of low power consumption and moderately high to
high throughput makes analog decoders practical and efficient for WSNs.

Time Synchronization Issues
• Sensor nodes are self-organizing and possess an important characteristic
that they synchronize among themselves to communicate with each other
(energy-efficient radio schedule).
• Furthermore, synchronization is also required for in-network processing,
acoustic ranging, distributing an acoustic beam forming, and for developing
other protocols designs/applications which require accurate time.
• Time synchronization in WSNs calls for special consideration due to their
limited energy, computation power, memory, and size of sensor nodes.
• In traditional networks as there is no such limitations, we have more
efficient solutions like GPS and network time protocol (NTP)
• In case of WSNs, time synchronization handles clock synchronization of
different sensor nodes in a single-hop and multihop environment.

• In WSNs, the prime concern is increasing the lifetime of network by energy
conservation, the operations like time synchronization should be efficient.
• Limited bandwidth restricts the data rate and hence, frequent messages
among the sensor nodes cannot be exchanged, which is the main
requirement of synchronization algorithms.
• Hardware limitations of sensor nodes further limit the processing capability
and the memory required for the storage of synchronization algorithms.

• To prevent collisions in time-division multiple access (TDMA) based
applications, sensor nodes should have synchronized clocks, so that the
noncommunicating nodes switch to sleep mode to conserve energy.
• In cross-layer network management, solutions such as intertwined medium
access scheduling and in-network data aggregating synchronized clocks are
required.
• Due to these resource constraints, the clock synchronization protocol
should be lightweight and efficient in terms of communication overhead.

Time Synchronization Issues: Time Synchronization Protocols
• The time synchronization protocols can be classified into three categories:
 Sender-receiver protocols
 Receiver-receiver protocols
 Receiver-only protocols

• Sender-receiver protocols:
 Sender–receiver protocol follows the following steps:
 Examples of sender–receiver protocols:
• Transmitting node sends periodic messages with its time stamp containing the local time.
• The receiver synchronizes its clock with the sender’s time stamp message.
• The delay message between sender and receiver is calculated by measuring the total time of
sending and receiving the message.
• The time sync protocol for sensor networks (TPSNs) (Ganeriwal et al.2003)
• Flooding time synchronization protocol (FTSP) (Maróti et al. 2004)
• Gradient time synchronization protocol (Sommer and Wattenhofer 2009)

• Receiver–Receiver-Based Protocols
• Receiver–only Protocols
 Achieved at a local level in contrast to synchronization achieved at network level by
some of the protocols.
 Example of receiver–receiver protocol:
• Reference broadcast synchronization (RBS) protocol: In this protocol, neighbors receive a
synchronization message by the node and this is used as a reference time to adjust their clocks
 A group of nodes do not send or receive the synchronization message, instead they
overhear the synchronization messages exchanged between a pair of sender–
receiver nodes working on the principle of sender–receiver protocol
 Example: Pairwise broadcast synchronization (PBS) (Kyoung-lae Noh et al. 2008).

Naming and Addressing Issues
• Naming and addressing issues are related to network management in WSNs
• Under this combined scheme, each node gets and identifies its and the
neighboring node’s name and location.
• Two basic approaches to naming are low level and high level.
• Application independent naming is low-level whereas high-level naming is
location independent.
• Applied when communication between applications is required.
• Unique node identifier (UID) provides a unique name to every node which
consists of such components as name of the supplier, name of the item, its
sequencing, etc

• A name to a component can be assigned at manufacturing time or a
temporary name can be assigned to increase the energy efficiency.
• The network identifier is used to distinguish the networks, which are
working into similar environment and geographical area.
• Message authentication code (MAC) address is used to differentiate
between neighbors of a node.
• Sometimes, a network address is essential to locate a node in multiple hop
scenarios which is related to routing.

• Uniqueness of addresses is classified as globally unique address, network-
wide unique address, and locally unique address.
• The globally unique address occurs at most once all over the world and
uses 48-bits MAC address.
• The network-wide unique address is unique only within a network.
• Local address can be used multiple times within a network.
• Naming schemes save energy when a node offers data with high attributes.
• Through naming:
 Useful information is passed to the neighboring nodes and thus the entire network
which reduces the overhead and latency.
 A sink node sends a query to the nodes.
 Queries are compared with the knowledge of attributes in a node.
 Then, the nodes pass answers to the sink node queries.

Naming and Addressing Issues: Address Allocation and Assignment
• In WSNs, dynamic address assignment protocols are used to allocate the
addresses a priori or on demand.
• In centralized address assignment scheme, a single authority node can
control and monitor the address pool.
• It has some limitations; the central node is reachable only until the network
is partitioned.
• If a node joins a group after the networks is partitioned, it cannot connect
to the central node.
• This approach is not suitable and creates significant traffic.

Naming and Addressing Issues: Address Allocation and Assignment
• In distributed WSNs, all nodes within a network can provide and accept the
same address assignment scheme.
• The node addresses need not to be unique all the time implying there may
be duplicity within the network and is called an address conflict.
• Therefore, research to resolve the address conflict detection and correction
is required.
• A distinction between strong and weak duplicate address detection (DAD)
is required.

Naming and Addressing Issues: Types of Addressing
• Application requirements of WSNs pose new challenges in node addressing.
• Fixed and universal addressing of sensor nodes is not a viable option
• Content-based and geographic addressing can be used for addressing as
shown below:

• Content-Based Addressing:
 This type of addressing is data-centric instead of id-centric.
 The data contents define the addressing instead of nodes.
 The middleware systems perform this type of addressing.
 In WSNs, the sensors sense data continuously and then the data of interest is used to
describe the addressing.

• Geographic Addressing:
 Idea is to use the location information available about a node locally for routing, i.e.,
its own location and that of its neighbors without knowledge of the entire network
 Ahmed (2012) applied a software- and hardware-based addressing scheme for WSNs.
 A scheme for long thin WSNs was proposed by Pan and Tseng (2012).
• The addressing and routing scheme was based on ZigBee protocol.
• A distributed address scheme is applied for the assignment of network address.
• Before address assignment, the users collect some information related to router and network:
- What is the maximum number of children of a router?
- What is the maximum number of child routers of a router? (It is limited to 5)
- How much is the depth of the network?
• The addresses are assigned in a systematic fashion from top to bottom.

• Research Issues and Challenges Related to Naming and Addressing
 The two main aspects in addressing are address assignment and address
representation. When the assignment is static, there is no true scaling issue. Dynamic
assignment needs to be explored.
 The addresses and names in a sensor networks can be used for nodes, MAC address,
network address, and network identifier. Research is ongoing on the energy-efficient
naming and addressing schemes.
 The issue occurs when the unique node ID, which is allocated before deployment, is
used as the MAC address.
 Research on geographic routing addresses two issues, one is routing packets
successfully in a given topology and second is acquiring location information of nodes
reflecting the given topology.

• Research Issues and Challenges Related to Naming and Addressing
 Spatial reuse of addresses requires a dynamic address assignment protocol. Such a
protocol can be centralized or distributed, but only distributed versions scale well.
Network wide unique addresses scale poorly. Spatial reuse dramatically improves the
scalability, as it is mainly the local node density and not the network size, which
dictates the address size.
 The problem of the malfunctioning occurs in software-based addressing. Resolving
this problem makes the sensor network energy inefficient.
 Although the unique and clear addressing is obtained with hardware-based scheme,
only the efficient design of the addressing scheme can ensure the elimination of
processing and transmission overheads.

Cross-Layer Issues
• Efficient utilization of the sensor network energy is the most effective way
to enhance its lifetime.
• Cross-layer techniques provide solutions to the load balancing, congestion,
bandwidth allocation, routing, transmission power, modulation, reliability,
data aggregation, packet overhead, and end-to-end delay.
• The cross-layer design on multiplr layers of the network protocol stack
reduces the chances of design improvements and level of modularity.

Cross-Layer Issues: Cross-Layer Interaction in Network Protocol Stack
• Combination and interaction between layers such as physical (PHY), MAC,
routing (ROUTE), and application (APP) gives the best solution to the
problem of energy efficiency and performance control.
• Cross-layer designs can be distributed, centralized, manager-based, or
nonmanager-based.
• Main cross-layer optimization techniques are classified according to the
design on the layers of network stack in WSNs e.g.
 PHY-MAC-ROUTE (PMR) approach
 PHY-MAC-APP (PMA) approach
 MAC-ROUTE (MR) approach
 PHY-MAC-ROUTE-APP (PMRA) approach,

Cross-Layer Issues: Open Research Challenges in Cross-Layer Design
• Signal fading and path loss effects should be considered while designing the
cross-layer protocols.
• Limited work has been done on the mobility effects on cross-layer design.
Thus, the mobility effects such as topology reconfigurations should also be
considered to fix the lower bound on energy.
• Cross-layer protocols, which reduce the path loss, congestion, and end-to-
end delay within network must be developed.
• Cross-layer protocols share and interchange large amount of data between
layers. This require large memory space, which is a limitation for WSNs.

Cross-Layer Issues: Open Research Challenges in Cross-Layer Design
• A design change in one component can affect the entire system which leads
to negative consequences such as the instability and modularity issues.
• The available discrete event simulators for WSNs, e.g., J-Sim, GloMoSim,
QualNet, OPNET, etc., work on traditional layered architecture. Therefore,
development of new software simulators for cross-layer implementation is
a big research challenge today.
• The cross-layer design issues are not limited to only these above-
mentioned points. A global optimum solution is required to minimize
energy consumption and maximize network performance.

Module 11
Data Aggregation in Wireless Sensor Networks

Contents
1. Elements of data aggregation
2. Energy-efficient data aggregation techniques
3. Security in data aggregation
4. Privacy preserving data aggregation

1. Understand the importance of data aggregation in WSNs.
2. Discuss energy-efficient data aggregation solutions in detail.

References
Chapter 7
Energy-Efficient Wireless Sensor Networks, Edited by:
Vidushi Sharma and Anuradha Pughat, CRC Press, 2018
(ISBN: 13: 978-1-4987-8334-7)

Introduction
• Usually, sensor networks are densely deployed to cover vast land spans or
geographical areas of interest.
• Primary target for WSNs is to gather data and provide information about the
environment to the neighboring nodes.
• Neighboring nodes may generate highly correlated and redundant data with
a focus on event detection application.
• This data is huge and, sometimes, the same events are likely to be gathered
and transmitted by other nodes too.
• Hence, wherever a gigantic amount of data is to be produced or processed,
data aggregation needs to be associated with the system.

Introduction
• In WSNs, data aggregation is required because of two main reasons.
• Studies have revealed that communication is more energy consuming than
computation i.e., the transmission cost of single bit information is thousand
times more as compared to energy spent in executing one instruction
• Main idea is to perform in-network processing to reduce communication
cost so that data aggregation is more valuable than reading, collecting, and
communicating raw sensor data (in spite of processing cost)
1. First reason is the huge amount of data produced by each node and large number of
nodes in each network, which gives a stack of data to process and analyze. This data
is to be converted into information relevant and valuable to the data consumer.
2. The second reason is the energy optimization in WSNs, as this has been too viable in
recent years. Data aggregation reduces the amount of transmission and processing
and, thus, the energy use.

Introduction
• A simple example of a sensor network with three nodes presenting data
aggregation is shown in Figure 7.1.
 Two nodes are taken as source nodes (Sr1
and Sr2) and the third as sink node (Si).
 Assume that d is the distance between two
nodes and it is too low.
 The nodes Sr1 and Sr2 are deployed to gather
similar data and send that to Si.
 The node Si consumes twice the energy in
receiving and processing similar data from
two source nodes.

Introduction
• Another example includes one more node, i.e., node C, which works as an
aggregator and only receives data from other two nodes and sends it to the
sink (see Figure 7.2)
 Here, the nodes Sr1 and Sr2 are gathering
the data and sending too, and they are
transmitting it over half of the distance
than in the prior case (Figure 7.1).
 This explains the importance of data
aggregation in WSNs.

Introduction
• Data aggregation in WSNs is defined as the process of gathering data from
multiple sensors on intermediate nodes using SQL queries or mathematical
functions in order to eliminate redundant transmission and provide fused
information only to the base station.
• The outcome of data aggregation on sensor network are reduced network
traffic, reduced energy consumption, and enhanced lifetime.

Elements of Data Aggregation
• Some performance parameters for energy-efficient data aggregation are:
• Accuracy of data
• Reliability
• Correlation coefficient
• Detection of false alarm
• Data redundancy
• Latency
• Power consumption
• Lifetime of the network

• Aggregation depends upon following elements, which effect its outcome:
 Network architecture
 Aggregation function
 Data representation
 Aggregation resources.

• The power consumption depends on:
• If the number of messages is low, there will be lesser energy consumed by
sensor networks at the cost of accuracy.
 Data aggregation elements
 Network time period
 Latency
 Data accuracy.

• In a data-centric approach, data aggregation is more energy efficient than
other approaches.

Energy-Efficient Data Aggregation Techniques
• One possible way of improving the reliability of WSNs is to deploy a few
redundant data nodes in the sensing area.
• The redundant nodes sense similar data and forward that data to the sink
nodes.
• The additional sensor nodes reduce the chances of network failure at the
cost of extra bandwidth use nd energy consumption in communicating and
processing redundant data.
• There exist a trade-off between reliability and energy consumption.

Energy-Efficient Data Aggregation Techniques
• Another way is to deploy a few nodes in the routing path, which performs
the energy-efficient data aggregation of the sensed data from its neighbors
and transmits the reduced data.
• This reduces redundancy and a huge amount of power consumption.
• Data aggregation approaches include:
 Centralized data aggregation
 In-network data aggregation
 Tree-based data aggregation
 Cluster-based data aggregation

Energy-Efficient Data Aggregation: Centralized Approach
• Better known as an address-centric approach because every node transmits
packets to a central node by a shortest possible path via multihop protocol.
• Each wireless node that captures data broadcast those packets to the
leader, which does not have power as the primary concern.
• The leader now processes data with aggregation queries, so that redundant
data can be eliminated.
• Other nodes are going to receive the data packet, but because they are
being addressed only to the centralized node, they won’t be processed.
• The packets will just get passed on to the next and to the next till they
reach the designated address.

Energy-Efficient Data Aggregation: In-Network Aggregation
• In-network aggregation is a decentralized approach.
• In this approach, instead of data processing through a central node, each
intermediate node performs the same task but on a smaller scale.
• The in-network aggregation scheme does not provide any centralized
processing facility (CPF).
• Each intermediate node functions as an independent node in the network
• There is no acknowledgment, i.e., it is one-way communication.
• The basic objective of reducing resource consumption (in particular,
energy) is fulfilled; therefore, network lifetime increases.

• A generalized aggregation, i.e., tiny aggregation (TAG), approach is
developed specifically for TinyOS-based sensor motes.
• It uses a declarative interface to collect and aggregate the data and
distributed aggregation queries to make the network energy efficient.
• The sensor data flows upward in the tree (from node leaf to the parent) to
reach the user.
• The in-network TAG approach reduces the number of packets transmitted,
reduces latency, and enhances the network lifetime.
• There are two ways for in-network aggregation:
 In-network aggregation with size reduction
 In-network aggregation without size reduction.

• In-network aggregation with size reduction:
 Aggregation with size reduction refers to the process of a node gathering data and
transmitting it without adding any address to the packets.
 When a packet is received by a neighboring node, that node combines it with the
data it has and compresses the data packet so that the packet length gets reduced
and it can be transmitted or forwarded toward the sink.

• In-network aggregation with size reduction:
 In-network data aggregation without size reduction refers to a node merging data
packets received from multiple neighbors.
 Instead of calculating the length size or compressing the data, the data packets are
transmitted directly.

Energy-Efficient Data Aggregation: Tree-Based Approach
• The in-network and centralized approaches are the broadcast approaches
having no predefined structure, while a tree-based sensor network has a
spanning tree like structure.
• The branches or leaves of the tree are considered as nodes and the sink is
defined as the root of the spanning tree.
• The network works in a similar manner as one will get to the root of a
spanning tree, starting from the top, i.e., leaves or nodes, and each node
will have a parent node where the data is to be transmitted.
• This flow keeps on running through all the nodes to the sink or root.
• Aggregation is carried out as we move to the roots of the spanning tree.

Energy-Efficient Data Aggregation: Tree-Based Approach
• On the basis of node synchronization, tree-based aggregation is divided
into two types:
 Synchronous tree-based aggregation
 Asynchronous tree-based aggregation.

Energy-Efficient Data Aggregation: Cluster-Based Data Aggregation
• In flat networks, nodes in a communication path perform data aggregation
while cluster heads perform data aggregation in hierarchical networks
• In comparison to flat networks, hierarchical networks suffer lowest latency
in transmission and are considered more reliable.
• Cluster-based aggregation approach refers to sensor networks where
clustering is done for data communication via single hop or multihop.
• Clustering in the sensor networks is defined in a schema, where the entire
network is divided into smaller networks or groups, called clusters.
• The clusters are divided according to locations of the nodes where
neighboring nodes form a cluster and within a cluster the node that is
either most power efficient or nearest to sink is selected as cluster head

Energy-Efficient Data Aggregation: Cluster-Based Data Aggregation
• In a cluster-based network, the mode of data communication is unicast.
• Nodes gather data and transmit it to the cluster head.
• Cluster head then does the processing and applies queries in order to
perform aggregation over the cluster.
• Many cluster-based data aggregation protocols have been introduced in
past few years
• This approach is more energy efficient, highly accurate and incur least
overhead.
• Furthermore, cluster heads can also play the role of data aggregator. Many
researchers have introduced this feature to the protocols for WSNs.

Security in Data Aggregation
• Data aggregation is performed with the help of an aggregation function
that takes raw gathered data from sensors as input and produce a fused
formed data as output.
• In order to secure data aggregation in sensor networks a secure data
aggregation function is required.
• One of the reason why security is essential in data aggregation is the
falsification of end results after data aggregation since a single malicious
node is capable of biasing entire aggregation results.
• Data privacy can be breached if a malicious node performs a man in the
middle attack, sniffs data, and forwards it to the sink too.
• This will also be a breach of confidentiality of data.

Security in Data Aggregation
• Different ways to secure the processing of aggregation from producing
falsified digest data include cryptography, quantiles aggregation, RANBAR,
voting, and verification.
• All of the above processes can be classified into three phases:
• Query diffusion
• Aggregation
• Verification

Security in Data Aggregation: Query Diffusion
• Query diffusion is the first phase in which the base station broadcasts SQL
query inside the network.
• The SQL query passes through various nodes.
• Typically, this process is energy-consuming because there is high possibility
that a node gets same query twice or thrice depending upon its location.
• Therefore, the query diffusion phase is combined with localization of the
nodes.
• Note that simple queries are insecure. Therefore, the complex query must
be used in order to construct a secure data aggregation structure

Security in Data Aggregation: Aggregation
• In aggregation phase, nodes (that fulfill a criteria of the discriminated
query) form clusters or spanning tree depending upon the query.
• For example:
• Assuming tree-based aggregation, the query diffusion takes place from the sink
node to very last leaf nodes.
• The acknowledgment will form a spanning tree, which will have a parent node and a
leaf node as per the query.
• If the distance between parent and leaf node is greater than the distance between
leaf nodes and the sink, the leaf node would not be part of the spanning tree.
• Instead, the leaf node will directly send data to sink node instead of parent node.
• However, this becomes more energy consuming, whereas the primary objective of
the aggregation is to make the sensor network energy efficient.
• Hence, a new approach which uses an aggregator as well as a forwarder is used.

Security in Data Aggregation: Verification
• In order to ensure secure network data flow, the base station must verify
every aggregated data from all the nodes
• Verification can be accomplished in multiple ways:
1. Using a sink that can verify every node, which sends the data to the aggregator.
2. Verify data aggregated at the aggregators only, at the cost of energy.

Privacy Preserving Data Aggregation
• Reasons due to which data privacy with aggregation is a necessity:
 Misuse of data is the most critical reason behind the privacy preservation, for
example, health monitoring applications, where a hospital periodically monitors the
patients’ various health indicators like blood pressure, sugar levels, etc. and stores
at database (sink). This becomes exclusively private information, which a hospital
has to keep confidential except from doctor or patient.
 Location of army vehicles in a battlefield should be kept private.
 Disclosing incompetence is another reason behind making the datasets private
 It is critical because the companies, government agencies, or any other public,
private, or nongovernment organizations are not allowed to violate the privacy of
any country. Legislation bounds them to preserve the privacy of users

• Two concerns, associated with privacy preservation are:
• PPDA is basically categorized into two types of protocols:
• Internal: Solution of maintaining internal privacy lies in securing the network and
making all the nodes trusted.
• External: To maintain external privacy, privacy-preserving data aggregation (PPDA) is
implied
• Homogeneous protocols: If all the nodes have the very same resource, then
homogeneous protocols are applied;
• Heterogeneous protocols: If there is more than one type of nodes in the network
such as aggregator nodes and leaf nodes, then heterogeneous protocols are applied.

• Furthermore, the protocols are divided into two types:
• The protocols are basically of three types used over various kinds of
unicast- and broadcast-based networks.
• These are perturbation, shuffling, and privacy homomorphism
 End-to-end encryption: The entire communication is encrypted. Apart from the sink
and node, no one can decrypt the packets making aggregation difficult to perform,
but decreases communication overhead and guarantees privacy preservation
 Hop-by-hop encryption: Sensor sends encrypted data packets to aggregators and
aggregator decrypts it; aggregates data, re-encrypts the aggregated data, and sends
it to the sink.

• Furthermore, the protocols are divided into two types:
• The protocols can generally be categorized into three types:
 End-to-end encryption: The entire communication is encrypted. Apart from the sink
and node, no one can decrypt the packets making aggregation difficult to perform,
but decreases communication overhead and guarantees privacy preservation
 Hop-by-hop encryption: Sensor sends encrypted data packets to aggregators and
aggregator decrypts it; aggregates data, re-encrypts the aggregated data, and sends
it to the sink.
 Perturbation
 Shuffling
 Privacy homomorphism

Privacy Preserving Data Aggregation: Perturbation
• In privacy preservation, sensor network always has an order or hierarchy in
a scenario where only the sink knows the network formation.
• Knowledge of data flow cannot help recognizing the route to an aggregator
or to the sink, and the network is called a perturbed sensor network.
• If every node divides its data into a polynomial of order k – 1, and k number
of nodes sends it to all other nodes using shared key, then a sensor sums all
the received polynomial and sends it up to the next aggregation level.
• The aggregator inverses the matrix and resolves it into separate packets
and aggregates them without knowing source of the packets and forward
them level up to the sink.
• Perturbation is an effective privacy-preserving technique, but it increases
the calculation overhead.

Privacy Preserving Data Aggregation: Shuffling
• In shuffling, the data is sliced into number of members of the nodes falling
under the aggregator; then that node keeps one slice and distributes the
other slices after encrypting with a private key to the rest of the nodes.
• With shuffling, exact origin of data cannot be recognized and even if an
aggregator is malicious, it will also receive encrypted packets in parts.
• Shuffling improves the privacy preservation, however, energy efficiency is
affected because of multiple hops.

Privacy Preserving Data Aggregation: Privacy Homomorphism
• An energy-efficient privacy preservation technique because of arithmetic
operations done on encrypted data and decryption process not required
• In privacy homomorphism, every leaf node shares a separate key with the
sink, but at the same time function adds a message with a key.
• This message when added up forms a key at the aggregator, which is used
to aggregate the data by summing up and finally the aggregator forward
the message to the sink.
• An effective technique for privacy preservation and energy efficiency, but
not scalable as the sink has to keep the different key of the nodes.

Challenges in Data Aggregation
• Network lifetime is one of the primary concerns of sensor networks and to
enhance it a systematic study of the relation between energy efficiency and
system lifetime must be conducted.
• Another area worth exploring is by analyzing the limits of the lifetimes of
sensor networks.
• Another potential area is to generalize sensor networks for data
aggregation in telecommunications by working on the mobility factor of
sensor networks.
• Security is an eminent issue in data aggregation applications. Integrating
security as an essential component of data aggregation protocols is one of
the interesting problems for future research.

Challenges in Data Aggregation
• Data aggregation in dynamic environments serves various challenges and
forms another candidate of future research work.
• Another interesting domain of research is the application of source coding
theory for data-gathering networks. Power savings in data aggregation
become crucial depending upon the fact that WSNs have resource
constraints.

Module 12
Sensor Network Security

Contents
1. Overview of security aspects in WSNs.
2. Vulnerability of WSNs to threats and attacks
3. Attacks in WSNs
4. Security mechanisms.
5. Cryptography.
6. Key management.

Contents
7. Authentication and integrity in WSNs.
8. Secure routing
9. Secure location
10. Secure data aggregation.

1. Describe security aspects in WSNs.
2. Explain various attacks in WSNs.
3. Understand security mechanisms to detect, prevent, and
recover from the security breaches.
4. Discuss the authentication and integrity in WSNs.
5. Describe the notion of secure routing and data aggregation.

Required Reading
1. Chapter 8 Energy-Efficient Wireless Sensor Networks,
Edited by: Vidushi Sharma and Anuradha Pughat, CRC
Press, 2018 (ISBN: 13: 978-1-4987-8334-7)
Recommended Reading
1. Chapter 12: Wireless Sensor Networks: A Networking Perspective, Jun
Zheng and Abbas Jamalipour, Wiley-IEEE Press, 2009. (ISBN: 978-0-470-
16763-2)
This Presentation is mainly dependent on the textbook: Energy-Efficient Wireless Sensor Networks, Edited by: Vidushi Sharma
and Anuradha Pughat, CRC Press

• Overview of security aspects in WSNs.

Security Goals
• WSNs have the following security goals:
Confidentiality
Integrity
Availability
Access control
Data origin and entity authentication
Nonrepudiation
Authorization
Privacy
Freshness

Security Goals
Forward secrecy:-
A node must not be endorsed to receive or send or know the messages which will be
transmitted in the future in the network after it disintegrates from the network.
Backward secrecy:-
A newly associated sensor node must not have access to any messages earlier sent on the
network.
• WSNs have performance-specific requirements based on their areas of applications as
follows:
 Self-organizing
 Scalability
 Time synchronization
 Efficiency
 Survivability

Security Goals
• Performance Metrics
Resilience
Resistance
Flexibility
Robustness
Assurance
• Security Limitations in WSNs
Limited resources
Unreliable communication
Unattended operation

Objective of Security
• To minimize the resource consumption and maximize the level of
security performance
• To identify security attacks on the WSN channel, including passive as
well as active interference
• To evolve less complex security schemes best suited for wireless
communication, sensor network.
• To develop security schemes which are able to handle increased
complexity as a result of large-scale deployment and node mobility
• To propose a security scheme to manage dynamic topology of the
network in view of node addition and node wearing out

Vulnerable Components
Base Station Security:
• A BS enables a WSN to communicate all the processed information to the outer
world via wireless medium.
• BS has more computational and communication capabilities
• More resilient to malicious activities like security breaches and attacks.
• The traditional security mechanisms for WSNs consider that the BS is secure and
robust as compared to other nodes in the network.
• If the adversary has more powerful and capable devices to breach security, the BS
may become a failure point.
• Deployment of multiple BSs to administer resistance against individual BS failure

Protect the BS : Method – 1
• Encrypting the packets and address field using pairwise shared key between
two neighbouring sensor nodes – hide the identity of BS
• Construct anonyms of the nodes using hash function – to hide the node IDs
• Anonyms of the nodes : source addresses or destination addresses.
• The generation and distribution of the pairwise shared keys are done by the
BS during the network topology formation phase.

Protect the BS : Method – 1
• An attacker needs to know the place where the BS has been installed in order
to launch an attack.
• Relocation of the BS so that location tracking of the BS is difficult for the
attacker.
• The attacker can attain the objective of finding the location of the BS by
analyzing traffic of the network. Henceforth, it is very essential to obscure
the traffic flow pattern and routes.
• To prevent this three methods are proposed.
In the first scheme, a multi-hop path is selected by the originating node for each data
flow which renders an attacker unfamiliar with the path from which traffic may flow.
The second scheme suggests the random creation of fake paths to confuse attackers.
In the last scheme, multiple random areas are designed or created for communication
activity to conceal the real location of the BS in the network.

Sensor Node Security:
• A sensor node is the smallest unit of WSNs which has low cost, low power,
and low storage space.
• These nodes notice the occurrence of any real-world event of interest and
process and communicate that information to the next node or level to be
available to the end user for predetermined purposes such as healthcare,
defence, monitoring, etc.
• The resource limitations of these nodes may result in security breaches or
attacks such as node capture, node réplication, etc.
• The attacks can be performed from outside or inside of the network and
categorized as external attacks and internal attacks.

• The external or outsider attack is performed when an unauthorized
node which is not a legitimate member of the network attacks it.
External have two categories: passive and active.
Passive attacks deal with unauthorized monitoring or “listening” to the information of
packets in the channel.
Active attacks, which are performed externally, interrupt working on the network by
intercepting the communication channel, fabrication, or replay of data packets, denial-
of-service (DoS) attack, jamming, etc.
• Internal or insider attack is performed when the attacker node is from
the legitimate nodes. An adversary can perform an internal attack by
compromising the sensor node.

Attacks in WSNs
Adversary's Capability-Based Attacks:
• An adversary, who may try to eavesdrop the wireless medium without directly
affecting the information transmitted is called a “passive attacker.”
• And an adversary called an “active attacker” may try to delay, replay, or inject
fabricated messages in the original data stream.
• The attack can be performed from outside or inside the network as per the
origination of attack.
• The attacks that are sourced from sensors, which are not part of the WSN, are
outsider attacks and insider attacks are performed by genuine sensors, which got
compromised.

Attacks in WSNs
Information in Transit-Based Attacks:
• In WSNs, sensors examine the occurrence of an event and subsequent changes in
parameters and other values and forward them toward the sink.
• The information sent to the sink is the processed report of the network activity,
which should reach the sink correctly and completely.
• This information if stolen by an attacker can be misused to gain unwanted advantage
and compromise the nodes in the network.
• The aim of the adversary is to disseminate false information and deceive network
users.
• The attacker can attack the information traveling on the network and can perform
replay attack, DoS, etc.

Attacks in WSNs
Host-Based Attacks:
• The system or the entity using that system may be corrupted and behave in
an unexpected manner.
• This attack can be divided into user compromise, hardware compromise, and
software compromise.
In the case of user compromise, the entities that are accessing the network are misled
and made to reveal the credentials, key material, etc.
Hardware compromise is done by the attacker when he either tampers with the
hardware machinery of the node or captures the node itself to destroy the node and
information stored in it.
In software compromise, the adversary tries to intrude in the network to access the 0
node’s software running inside it to launch a malicious attack.

Attacks in WSNs
Network-Based Attacks:
• The network-based attacks are the attacks which are being performed on the
communication layers of the network protocol.
• Such attacks may be performed from inside or outside the network in which
sometimes the attacker does not want to cause direct loss like modification,
destruction, or fabrication of information but wants to access the network for
his own advantage.
• Based on the network/protocol stack, the layer wise problems are as follows.
Physical Layer Attacks
 Jamming
 Tampering or destruction
 Radio interference

Attacks in WSNs
Data Link Layer Attacks:
• This layer has the function of transmitting data on a physical link and provides
networking media.
• It is generally associated with network access, topology, packet delivery, flow control,
etc.
• So, on this layer the attacker can perform attacks to interrupt these functions of the data
link layer and those attacks may include the following:
 Collision
 Continuous channel access or exhaustion
 Unfairness

Attacks in WSNs
Network Layer Attacks:
• This communication layer is responsible for addressing, routing, and end-to-
end delivery of the packets. So, the attacker can create problems in routing or
packet delivery, etc., and affect the network operation and security through
following the attacks:
Sinkhole
Hello flood
Selective forwarding/black hole attack/neglect and greed
Node capture
Wormhole attack
Spoofed, altered, or replayed routing information

Attacks in WSNs
Transport Layer Attacks:
• This layer of the communication protocol stack helps in transmission of the
packet to the destination and reassembling them. To disturb the data delivery
and create vagueness, the attacker can perform the following attacks:
Desynchronization attack
Flooding
Application Layer Attacks:
• This layer acts as the interface for different user applications and enables
access to the Internet.
• This layer introduces synchronization, data integrity, and error control. To
interrupt the interoperability and functionality of this layer, an attacker can
perform the following attacks:
Path-based DOS attack
Overwhelm attack

Security Mechanisms
• The motive of any security mechanisms is to detect, prevent, and
recover from the security breaches as and when they occur.
• To secure the WSNs effectively, the security mechanism must fulfil
certain criteria like resiliency, fault-tolerance, energy efficiency,
scalability, flexibility, self-healing, etc.
• A security scheme should propose choices in view of desired qualities
like reliability, latency, etc.
• There are several security mechanisms proposed by various
researchers to encounter attacks and other issues related to the
security in sensor networks.

Security Mechanisms
Attacks-Based Security Schemes:
• An attacker can launch an attack in a WSN as per his capability and resources.
On the basis of his ability and resource capability, he can launch attack from
either inside or outside the network with devices of similar or more
functionality. The WSN can be protected from such kind of attacks
• By deploying strong and robust security mechanisms. Such mechanisms are
designed in consideration of the constrained capabilities of WSNs.
• The attacker may have an intention to affect the network by simply
eavesdropping, compromising the nodes or taking over the whole network
itself.

Security Mechanisms
• When the attacker directly affects the network he can interrupt the network
operation, intercept the flowing traffic, inject, or fabricate the information in
traffic.
• These attacks can be prevented by implementing a security scheme which
may address intrusion detection and prevention, authentication, tamper
resistance, etc.
• In WSNs, an attack on information in transit results whenever any event
occurs and the sensors report it to the sink. The information being sent may
be compromised to supply false information to BSs or sinks. This may lead to
information interruption, interception, modification, fabrication, and
replaying.

Security Mechanisms
• The attacks by the intruder such as user compromise, hardware compromise,
and software compromise become the target for extracting vital information
in the WSN such as passwords, encryption and decryption keys, operating
system, and other communications facilitating information
• This leads to spoofing, node capture, and compromise, which further lead to
more compromised pairwise keys and therefore affect the security of the
network.
• Security attacks on the network layers target the information exchange
happening over the different protocols of the layer and during this
communication, the physical layer mostly suffers from jamming problem
which involves DOS attack.

Cryptography
• Cryptography is the study and art of encrypting the simple data or plaintext
and decrypting coded data or cipher text for security from adversary or
attack.
• For the protection of the sensor nodes from different security attacks such as
Sybil attack and blackhole attack and maintain data confidentiality and
integrity, there is a need for robust encryption techniques to be deployed in
the network.
• The encryption techniques may involve the deployment of both symmetric
key ciphers and asymmetric key ciphers.

Cryptography
Symmetric key system:-
• The sender and receiver share and use a common key that is saved
and kept secret from others.
• The sender encrypts a plaintext M with the key K by an encryption
algorithm E to get a cipher text C = E (M, K).
• At the receiving end, the cipher text C and the key K are given as input
by the receiver into a decryption algorithm D to get the original
readable plaintext M = D(C, K).

• Symmetric Key Cryptography
• Symmetric key cryptography uses the
same key for encryption and
decryption as shown in Figure 1.
• The symmetric key systems such as
Advanced Encryption Standard (AES)
(Daemen and Rijmen 2013), Data
Encryption Standard (DES) (FIPS PUB
1993), or Rivest Cipher 5 (RC5) (Rivest
1996), etc., is most logical and efficient
to be deployed in such limited
resources of the sensor network.
• This secret key system requires
scrambling or substitution operations,
hashing, rotation, or shifting, etc.,
which can be efficiently designed and
implemented in hardware or software.
Figure 1 Symmetric key cryptography

Cryptography
Asymmetric or Public Key Cryptography:
• Asymmetric or public key cryptography uses different keys to encrypt and
decrypt namely public and private keys as in Figure 2 a and b.
• Asymmetric key systems such as the Rivest–Shamir–Adleman (RSA) (Rivest et
al. 1978) algorithm, Diffie–Hellman key exchange (DHKE) (Diffie and Hellman
1976), digital signature standard, etc., are very secure and robust when
compared with the symmetric key system.
• The important cryptographic techniques are given in Figure 3.

Cryptography
Figure 2 Asymmetric key cryptography: (a) Encryption
with public key; (b) Encryption with
private key.
Figure 3 The main cryptographic scheme.

Key Management
• Cryptographic methods involve the use of keys (symmetric or
asymmetric) and these keys need to be handled carefully.
• The key distribution can be done in three ways
• randomly, predetermined (pre-distributed or stored in the node), and hybrid.
• key management is the process by which cryptographic keys are
generated, stored, protected, transferred, loaded, used, and
destroyed.
• The main objective of key management is to establish and maintain
secure channels among the communicating parties.

Key Management
• Key management schemes use keys for the secure and efficient
(re)distribution, and at times, generation of the secure channel
communication keys to the communicating parties.
• Communication of keys may be through pairwise keys, which are used
to secure a communication channel between two nodes that are in
direct or indirect communications or grouped keys shared by multiple
nodes.
• Network keys (both administrative and communication keys) may
need to be changed (rekeyed) to maintain secrecy and resilience to
attacks, failures, or network topology changes.

Key Management
Symmetric Key Management:
• Entity-Based Schemes
Entity-based schemes, which may be
called as arbitrated schemes are based
on trusted entity for key distributions
and key establishment.
• Master key-based pre-distribution
scheme
In this scheme, a master key is to be
pre-distributed and stored in each
sensor node of the network.
A random number and the decided
master key communicated within
nodes help to establish pairwise keys
between each sensor.
Figure 4 Symmetric key management schemes.

Key Management
BS participation scheme:
In this type of scheme, the BS plays a vital role in distribution of the keys to sensor
nodes. SPINS is a BS participation scheme, which enable each sensor to store a shared
key with the BS.
Whenever two sensors are required to communicate, a pairwise key can be sent by the
BS, which is encrypted with the shared key. This scheme provides resiliency, but not
scalability.
A trusted third node-based scheme:
This scheme relies on a common trusted third node for the key establishment between
two nodes.
This scheme provides resiliency and scalability as there is no need to store any master
keys in the node,

Key Management
Pairwise Key Pre-distribution Scheme:
As per the scheme, a pre-distributed key is stored in each node before
deploying the node in the WSN. This scheme offered good resiliency and
authentication, that is, even if one node is captured, the keys of other nodes
are safe.
Pure Probabilistic Key Pre-distribution Schemes:
• In this scheme, communication keys are established in three phases:
Key pre-distribution, shared-key discovery, and path-key establishment.
For key pre-distribution, a large pool of P keys is generated and then k distinct keys out of
P are drowned and loaded into each sensor node memory.
 The security of a communication depends upon the key connectivity for discovering the shared key and
establishing the path key

Key Management
Polynomial-Based Key Pre-distribution Schemes:
• In this scheme, the keys were distributed through polynomials. A pairwise key
pre-distribution using polynomial pools utilized a multiple random bivariate
polynomials.
• When the polynomial pool had only one polynomial, the general framework
degenerated into the polynomial-based key pre-distribution.
• The pairwise key establishment was carried out in three steps: setup, direct
key establishment, and path-key establishment. This scheme offered better
security and the scalability in WSNs.

Key Management
Matrix-Based Key Pre-distribution Schemes:
• A matrix-based group key pre-distribution scheme in which a symmetric
matrix Kn*n stores all pairwise keys of n nodes group, where the key of node i
represented by element kij for securing the link with node j. Each node i stores
the ith row of the secret matrix and the ith column of the public matrix G.
• After network deployment, each pair of nodes i and j can individually figure a
pairwise key kij = kji by exchanging their columns in plaintext as the key is the
product of their own row and the other’s column.
• Their rows are always kept secret. This scheme is said to be l-security means if
more than l rows are compromised, the entire secret matrix can be extracted
or broken by an attacker.

Key Management
Tree-Based Key Pre-distribution Schemes:
• This key management mechanism pre-distributed the keys to the nodes
arranged in a tree-like structure in the network.
• This can be further of two types: the star-like tree and binary logical tree.
Star-like tree-based key pre-distribution schemes
These schemes were based on strongly regular graphs and random graphs
correspondingly.
Logical tree-based key pre-distribution schemes
In this scheme, there were keys associated with a tree structure which is retained only by
one group controller, where every node corresponded to a key encryption key (KEK).
Every node as a component of the group communicates to a leaf of the tree and keeps a
node’s KEK from its leaf to tree roots. Then the root of the tree keeps the group key.

Key Management
Combinatorial Design-Based Key Pre-distribution Schemes:
• The combinatorial design theory was based on the existence and construction
of systems of finite sets whose intersections have specified numerical
properties.
Exclusion Basis System-Based Key Pre-distribution Schemes:
• The exclusion basis system (EBS) aiming at improved key management
efficiency and reduced overhead in group communications.
• EBS was based on combinatorial optimization methodology for key
management of group communication networks.

Key Management
Asymmetric Key Management Schemes:
• A public key cryptosystem is widely accepted and used for providing security
of data and networks in the realm of the Internet. RSA and ECC are two
major asymmetric cryptographic key techniques. It is worthy to note that the
security provided by a 160-bit ECC key is almost the same level of security
provided by 1024-bit RSA key.
RSA-Based Asymmetric Encryption System
RSA is a block cipher which uses two exponents e and d, where e is public and d is
private.
ECC-Based Asymmetric Encryption System
ECC is very efficient and is frequently used in the current network and data security
mechanisms.

Key Management
ID-Based Key Agreement Schemes:
• An IBE-based scheme promoted the usage of an arbitrary data (ID) which can be his/her
email id, etc., for computation of the user’s public key instead of taking it from a CA-
issued certificate.
Hybrid Schemes:
• These hybrid schemes utilize the merits of secret and public key systems which have
been mentioned in sections discussed above. In such schemes, it has been noticed that
the BS, sink, and cluster heads play an important role, being more resourceful as
compared to normal sensor nodes.
• They may be assigned the duty of performing some cryptographic computations and
broadcast the same to other sensors, as per requirement.
• The hybrid key establishment schemes reduce the high computational cost of the
sensors by placing them on the BS side. Such kinds of schemes are very efficient and
suitable for large-scale WSNs.

• Authentication and Integrity in WSNs

Authentication and Integrity in WSNs
• When two parties are in a communication protocol, each party
remains attentive for the legitimacy of the other party with whom
one is communicating.
• This can be achieved by authentication of the either parties through
some predefined means.
• The authentication and integrity can be accomplished by means of
suitable protocol or scheme, etc., deployed with the WSNs.
One-Hop Authentication:
In one-hop unicast authentication, each packet of the message is verified
between neighbouring nodes at the link layer by the link-layer key.

Authentication and Integrity in WSNs
Multi-hop Authentication:
• The link-layer authentication is considered not secure as the intermediate
nodes are not trustworthy and may modify the data payload. So, higher layer,
that is, transport layer authentication is required as it maintains end-to-end
connection.
• A multi-hop key can be computed between two end nodes via multi-hop path
while using a symmetric key system. This key negotiation may be unsuccessful
if any of the intermediate nodes on the path is compromised.
Broadcast Authentication:
• Broadcast is desired when some common message needs to be sent to a
group of nodes in WSNs. Each broadcast packet should be authenticated so
that attackers cannot inject false information. Symmetric key-based
techniques are assumed efficient, as they require a single secret key. A one-
way hash chain (OHC) is one of the basic techniques.

Secure Routing
• When a packet travels from source to destination, it is being routed through the
network and a path is formed till the destination.
• Routing protocols are the most important factors in deciding a path from the
source to the destination and finally delivery of data
• The path followed should be secure for preventing any data loss.
• If any attack occurs and affect routing protocols, high-layer applications also get
affected and the whole network may be compromised. So, it is necessary to
provide security at the routing level also for proper network functionality.
• Multipath routing can be used to overcome the selective forwarding attack to
increase the data delivery. Such techniques are good for preventing external
attacks, but internal attacks also needed to be countered.

Secure Location
Secure Location Scheme with Beacons:
• The schemes, which utilize beacons, have sensors with a positioning system
like the global positioning system (GPS) to find the whole location of the
network.
• The positioning and ranging mechanisms are designed to find location of the
node on the basis of measurements of the received signal strength and the
times of flight of radio or ultrasound signals.
Secure Location Scheme without Beacons:
• A beacon-less location discovery scheme assumes that sensors of the same
group are placed at the same time and at the same point and their locations
may have a probability distribution that can be investigated.

Secure Data Aggregation
• The process of data combining and aggregating may be called as data
fusion and data aggregation. The data aggregation is efficient in
reducing communication overhead.
Plaintext-Based Scheme
Plaintext-based schemes are methods in which data aggregation is done in a
readable form.
• Scheme Defending Against One Compromised Node
• Bidirectional Authentication Schemes
• Neighbour's Certificate Schemes
• Statistical Method

Secure Data Aggregation
Cipher-Based Scheme:
• In this cipher text-based scheme, the intermediary nodes along the path do
not have any knowledge of the transmitted data packet contents.
• The scheme, concealed data aggregation, obscures sensed data end-to-end
without affecting efficient in-network data aggregation.
• This scheme used an encryption transformation algorithm called a privacy
homomorphism (PH).

Main Reference
1. Chapter 8:Energy-Efficient Wireless Sensor Networks,
Edited by: Vidushi Sharma and Anuradha Pughat, CRC
Press, 2018 (ISBN: 13: 978-1-4987-8334-7)

Module 13
The Future for Sensor Networks – Cloud and IoT

Contents
1. Introduction.
2. Sensor Networks - Big Data & Cloud Computing to IoT
3. Web 3.0
4. IoT and IoE.
5. Fog Computing.
6. Infrastructure and data challenges - Collection, integration,
and analysis.

Contents
7. Data security and privacy
8. Distributed architecture to enable local actionable
Performance and scalability challenges
9. Intersection of AI, Data Science, and Machine Learning with
Sensor Networks’ data in IoT
10.Success factors for mass adoption and commercialization of
IoT

1. Discuss the intersection of sensor networks with Big Data,
Cloud Computing and Internet of Things (IoT).
2. Describe the concepts of data security and privacy.
3. Discuss emerging technologies and applications of WSN.

Required Reading
1. Chapter 11 Energy-Efficient Wireless Sensor Networks,
Edited by: Vidushi Sharma & Anuradha Pughat, CRC Press,
2018 (ISBN: 13: 978-1-4987-8334-7)
Recommended Reading
1. Chapter 18: Wireless Sensor Networks: A Networking Perspective, Jun
Zheng and Abbas Jamalipour, Wiley-IEEE Press, 2009. (ISBN: 978-0-470-
16763-2)

Introduction
• WSNs are only the means to an end, which is to make smart and intelligent
decisions based on the data collected.
• Sensor networks, Web 3.0, IoT and IoE.
• Fog computing Infrastructure and data challenges - Collection, integration,
and analysis.
• Data security and privacy and Distributed architecture to enable local
actionable intelligence and insights will also discussed.
• Performance and scalability challenges and Intersection of AI, Data Science,
and Machine Learning with Sensor Networks’ data in IoT will also study.
• Success factors for mass adoption and commercialization of IoT will also
define.

• Sensor Networks - From Big Data and Cloud
Computing to IoT.

Sensor Networks - From Big Data and Cloud Computing to IoT
Figure 1: Role of sensors in the next generation of applications including IoT.

• Sensors in drive of next generation applications - Figure 1
• Big data - very large data sets that cannot be processed using
traditional methods to draw intelligence from them.
• Context of applying Cloud computing
• Business application : enable their employees and customers to conduct
business operations.
• Individual application: to store files, documents, photographs, etc.,
• An easy solution to store once and access it anywhere where they are
connected to the Internet.

• Cloud computing services -SaaS, PaaS, and IaaS.
• Other Cloud Computing Services
• Database as a Service(DaaS)
• Storage as a Service (SaaS)
• Network as a Service (NaaS)
• Monitoring as a Service (MaaS)
• Communications as a Service (CaaS)
• Identity as a Service (IDaaS)
• Another classification of cloud computing involves
• Public cloud, Private cloud & Hybrid Cloud.

Web 3.0
• Web 3.0 can be expected to be more
• Connected
• Open
• Intelligent
• With semantic Web technologies
• Distributed databases
• Distributed computing
• Natural language processing
• Machine learning
• Machine reasoning
• Autonomous agents.

IoT and IoE
Figure2: IoT architecture layers.

IoT and IoE
• While the IoT helps in connecting physical objects via networks to capture,
transmit, store, and process data, the IoE adds people to the mix of
objects, data, and process.
• A key characteristic of IoT environments is
• The devices and objects can be quite heterogeneous
• Work under different physical conditions
• Generate data in different formats
• and be able to transmit this data using heterogeneous networks.
• Another key characteristic of the IoT is that
• Devices and objects can interact with each other without human intervention,
thereby relieving people to be available for other more valuable work.

IoT and IoE
• The ultimate goal of the IoT and IoE is
• To make life better for human beings by providing data-driven decision
making
• Actions without the need for human interference or with minimal human
intervention.

Fog Computing
• FOG COMPUTING- decentralization of data analysis and decision
making from the cloud computing architecture.
• Fog computing provides the low latency and also avoids the network
bandwidth limitations thus making the architecture more scalable,
robust, secure, and reliable.
• So while cloud computing is here to stay with all its advantages, fog
computing is emerging as a crucial supplement to the IoT architecture
due to its advantages of low latency, low network bandwidth
requirements, security, and reliability.

• Infrastructure and Data Challenges—Collection,
Integration, and Analysis

Infrastructure & Data Challenges-Collection, Integration, and Analysis
• One of the biggest challenges for the IoT is determining the right
architecture.
• The amount and type of data being generated, collected,
transmitted, stored, and processed are big consideration in the
design of an IoT system.
• Thus, scalability and performance are big challenges even when the
data is structured

Infrastructure & Data Challenges Collection, Integration, and Analysis
• When the data combines structured as well as unstructured data,
the problems of integrating this data becomes a challenge.
• Distributed processing and distributed databases have provided
architectural solutions to enable organizations deal with the large
amounts and variety of data.
• Another consideration is whether there is a requirement for any
pre-processing of data that needs to happen before sending the
data to the cloud.

• Another challenge for enterprises collecting and storing big data is
the building up of enough capacity to be able to back up all this
data.
• As data volumes grow, organizations will need to define retention
policies and automate selective backup of the data that must be
archived.
• This selection process of data to keep versus data to delete will add
to the workload increasing the need for processing, storage, and
network resources that may already be short.

• Device and sensor manufacturers as well as IoT platform providers
make available APIs. Some of the APIs are listed here.
• The BloomSky API
• The Fencer API
• AT&T M2X Keys API
• AT&T M2X REST API
• AT&T M2X MQTT API
• REST-like HTTP API
• SNAP PAC REST
• Predix Traffic Planning API
• The IOS tash IoT PaaS API

Infrastructure & Data Challenges Collection, Integration, and
Analysis
• The OGC Sensor Things REST API
• Netbeast API
• Kaa Admin API
• Insta Unite API
• Kontakt.io API
• The METAQRCODE API
• Weaver API
• Pimatic API
• Unification Engine API
• Sensorist API
• Web MIDI API

• However, the lack of standardization of such APIs as well as
specifications that provide a common way to describe the data is a
big challenge needed to be addressed if the IoT is to be successful.
• There are many efforts under way to address this standardization
problem.
• A successful IoT architecture would, therefore, stress and focus on
making data integration an integral part of the strategy.

Data Security and Privacy
Figure 3 Data flow diagram depicting flows of data from devices/sensors to IoT applications
on user devices.

• IoT cloud providers are very serious about data privacy and security
threats.
• They carry out what is termed a threat model analysis using data flow
diagrams, which detail out the various places where data originates,
transforms, transmits, or gets stored for further use. (See Figure 3)
• Mitigation techniques and technologies are then applied at each
point of the data flow diagram as well as to each data flow process
with the aim of establishing trust boundaries.

• Various threat categories like spoofing, tampering, repudiation, denial
of service, and elevation of privilege are considered at each data flow
node.
• There are many industry-standard techniques including cryptography,
multifactor authentication, identity management, trust management,
access control lists and permissions, as well as digital signatures and
identifiers.

• Distributed Architecture to Enable Local Actionable
Intelligence and Insights

Distributed Architecture to Enable Local Actionable Intelligence
and Insights
• In many use cases like parking lot management and road traffic
control in a smart city, it would be better to have geographically
localized data processing.
• In yet other cases like smart cars, it would just not be practical to
collect, transmit, and process data in a centralized database because
of the near real-time requirements of cars interacting with each other.
• This brought into focus different distributed architectures for IoT
where finding intelligence and actionable real-time inputs was
necessary.

and Insights
• A distributed IoT architecture for smart cars, thus, would be restricted to
specified geographic limits to be able to provide the required closed-loop
control as well as robustness needed for smart cars interoperability.
• Having such distributed architecture to support the IoT naturally makes
real-time data analytics a possibility.
• If local nodes could process specific data in real time, all devices and
sensors within that geographic zone could potentially benefit from real-
time actionable intelligence.

and Insights
• Such local nodes could simply be virtual machines supported by
appropriate data processing applications that could process raw data
to provide actionable intelligence.
• These localized data processing nodes would not just function stand-
alone but rather become a part of the larger IoT architecture.
• Such distributed local nodes can be used to transmit aggregated data
to the central IoT cloud databases.

• Performance and Scalability Challenges

Performance and Scalability Challenges
• The very nature of IoT systems makes performance, scalability,
availability, and resiliency a challenge.
• With the ever-growing number of devices and sensors connected to
any IoT system, it becomes a challenge to support heterogeneous
devices and data formats while at the same time being scalable.
• Thus, newer hardware and software architectures as well as
standards like EPCglobal (Electronic Product Code), which defines
how radio-frequency identification (RFID) data is collected, filtered,
aggregated, and transmitted, are being developed to address these
concerns.

• There are standards being created for the different layers of IoT
architecture:
• For the infrastructure layer- 6LowPAN, IPv6, RPL
• For identification layer- EPC, uCode, IPv6, URIs
• For the communication layer-WiFi, Bluetooth, low-power wide-area network
(LPWAN)
• For the discovery layer- Physical Web, multicast Domain Name System (mDNS), DNS
service discovery (DNS-SD);
• For the data exchange layer- MQTT, Constrained Application Protocol (CoAP), AMQP,
Websocket, Node;

• For device management layer like TR-069, Open Mobile Alliance-Device
Management (OMA-DM);
• For semantic layer like JSON-LD, Web Thing Model; and for the multilayer
framework layer like Alljoyn, IoTivity, Weave, Homekit, etc.
• To achieve higher performance and scalability in IoT architectures, more and
more processing capability needs to be pushed down to the nodes as well as
devices that are not limited by the constraints of power, computing, and
storage resources.

• Intersection of AI, Data Science, and Machine Learning with
Sensor Networks’ Data in IoT

Intersection of AI, Data Science, and Machine Learning with Sensor
Networks’ Data in IoT
• In recent years, though, there are two major factors that are shaping
the advancement and adoption of AI.
• One, the Internet and its myriad applications have started generating massive
amounts of data that is being stored in data centres around the world.
• Two, the low cost as well as more power of computers and storage has
significantly increased the amount of storage capacity as well as computing
power by orders of magnitude.
• The deployment of the vast amount of sensor devices makes the
collection, transmission, and storing of data easy for an IoT system.

Intersection of AI, Data Science, and Machine Learning with Sensor
Networks’ Data in IoT
• The availability of computing resources at the nodes available with
cloud computing uses machine learning, deep learning, data science,
and algorithms on this data to create intelligence.
• The more the data, the better the machine learning.
• The more the computation power, the shorter the time period to run
machine learning.

• Success Factors for Mass Adoption and Commercialization
of IoT

Success Factors for Mass Adoption and Commercialization of IoT
• We can categorize these success factors into two.
• The first category comprises of technology- and infrastructure-related factors.
• The second category comprises of business- and consumer-related factors.
Let us briefly consider each of these.
• On the technology side, the first big challenge is the interoperability
of devices and standards for many things like networks, data
transmission, etc. IoT systems need to have the devices and sensors
communicating with each other in the most efficient manner.
• The next technology challenge is connectivity and networks.

• With innovations like fog/node computing, this can be addressed to
some extent by using protocols like Wi-Fi to connect the local
devices/sensors to newly installed fog nodes for collation and
processing of data.
• This way, existing sensors only need to be connected to
geographically localized nodes which are further connected to the IoT
network.
• Other technology factors are automation, development of better AI
applications to provide actionable intelligence in real time.

• On the business and consumer side, the success factors include the
creation of an ecosystem whereby manufacturers and consumers
identify standard products that could be used for IoT systems.
• From a provider perspective, cost control while deploying a global-
scale IoT system can be a big challenge before specific consumer-
oriented applications could be offered to monetize it.
• To provide comprehensive solutions, exploring new business models
and engaging with a broader ecosystem are critical factors for
success.

Main Reference
1. Chapter 11:Energy-Efficient Wireless Sensor
Networks, Edited by: Vidushi Sharma and
Anuradha Pughat, CRC Press, 2018 (ISBN: 13:
978-1-4987-8334-7)

IT484 Cyber Forensics_Information Technology

More Related Content

Similar to IT484 Cyber Forensics_Information Technology (20)

Recently uploaded (20)

IT484 Cyber Forensics_Information Technology