IRJET- Competitive Analysis of Attacks on Social Media

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 03 | Mar-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1650
Competitive Analysis Of Attacks On Social Media
Narinderpal kaur1, Rasleen Kaur2
1Student, Dept of Computer Science Engineering, GIMET Amritsar, Punjab, India
2Assistant Professor, Dept of Computer Science Engineering, GIMET Amritsar, Punjab ,India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Today social media play very important role in
the communication process. With the help of social media
number of tasks is being accomplished. Normally socialmedia
is used to sell products, Career Consultancy, Communication,
sharing resources etc. Along with the advantages there are
number of disadvantages of the social mediaalso.Socialmedia
uses number of mechanisms to create users. And increase
their database. But they do not ensure validity of information
provided bsy the user. This will cause the deception. Deception
will cause number of problems. There are number of types of
deceptions which exist over the internet. Deception model is
prepared in order to analyze these problems. Some of the
deceptions are difficult to detect than the others. Some of the
challenges which social media must address are considered in
this paper.
Key Words: Deception, social media, Internet
1. INTRODUCTION
With the growth of the internet social media growth is
increased.[7] The social media is used for wide variety of
purposes. Social media is used to share user generated
contents to large number of other users. With that the
number of services provided by the social media also
increases. With the advent of technology the deception also
has been increased. Deception is caused due to falsifying
information provided by the user.[3] Social media provide
new environment for the deceivers to perform illegal tasks
over the internet. The main cause of deception is that it is
very easy to create account over the social media like
Facebook, Twitter etc. No verification of records is done in
case of the social media. They are just consider the increase
of database and do not consider deception.In this paper we
consider or attack deception as the deliberate attempt to
provide falsifying information to conduct harm over the
network.[31] The problem is extravagated since receiver
does not know about the deception. Because of which
receiver privacy will be on the stakes. The private
information of the receiver will be determined by the
deceiver. These false beliefs are transferred through verbal
and non verbal communications.[18] Clone attacks are
common source of deception over the social media. Rest of
the paper is focused on determining clone attack detection
strategies which result in redundant information or users
over the social media. Today social media assume critical
part in the correspondence procedure.Withtheassistanceof
social media number of errands is being refined. Ordinarily
social media is utilized to offer items, Career Consultancy,
Communication, sharing assets and so on. Alongside the
points of interest there are number of burdens of the social
media too. Social media utilizes number of components to
make clients. What's more, increment their database. In any
case, they don't guarantee legitimacy of data gave by the
client. This will cause the double dealing. Misdirection will
cause number of issues. There are number of kindsofdouble
dealings which exist over the web. Double dealing model is
set up keeping in mind the end goal to examine these issues.
A portionof the double dealingsare hard to distinguish than
the others. A portion of the difficulties which social media
must address are considered in this paper.
1.1 Machine Learning
To make the proposed counterfeit record identification
framework adaptable; we composed and executed a
commonsense machine learning pipeline including a
grouping of information pre-preparing; highlightextraction;
forecast and approval stages.[5] Machine learning
calculations that have assumed a noteworthy part in social
media examination incorporate Decision tree learning ,
NaïveBayes, Nearest Neighbor classifier, MaximumEntropy
strategy , Support vector machine(SVM) ,Dynamic Language
Model classifier , direct relapse and calculated relapse ,
Simple calculated classifier , Bayes Net and Multilayer
Perceptron.[18]The pipeline comprisesof threenoteworthy
segments; which we depict beneath and show in Figure.
Fig 1.1 Machine Learing
The Cluster Builder, as its name suggests, takes the crude
rundown of records and constructs groups of records
alongside their crude highlights.[15]The moduletakesclient
specialized parameters for (1) least and greatest group
estimate;(2) time traverse of records enlisted (e.g. most
recent 24 hours, a week ago), and (3) grouping criteria. The
grouping criteria can be asbasic as gathering allrecordsthat

offer a typical singe act eristic, for example, IP address, or a
more unpredictable grouping calculation, for example, k -
implies.[8] Once the underlying groups are manufactured,
client denied criteria can be added to later out a portion of
the groups that are not prone to be suspicious or may
present high false positives. For instance, one may wish to
later out records enrolled from the OSN's corporateIPspace,
as these are probably going to be test accounts that ought
not to be confined. The Cluster Builder takes crude part
genius le tables as info and yields a table of records with
highlights that are required for include designing, for
example, part's name, organization, and instruction. Each
line of the table speaks to one record and contains a “cluster
identifier" novel to that record's group.[28] This table is
utilized as contribution to the Problem Featurizer. In the
preparation stage the Cluster Builder should likewise utilize
account-level marks to name each group as genuine or
phony. While most groups have either all records or no
records named as phony, there will as a rule be a couple of
groups with a few records in each gathering. Along these
lines to process group labels, we pick an edge x to such an
extent that the groups with less than x percent counterfeit
records are named genuine and those with more prominent
than x percent counterfeit are marked phony.[3] The
operation optimal decision of x reliesuponexactness/review
tradeoffs (i.e., higher estimations of x increment accuracy to
the detriment of review).[27]
1.2 Profile Featurizer
The Profile Featurizer is the key segment of the pipeline. Its
motivation is to change over the crude information for each
cluster (i.e. the information for the greater part of the
individual records in the cluster) into a solitary numerical
vector speaking to the cluster that can be utilized as a partof
a machine learning calculation.[16] It is executed as an
arrangement of capacities intended to catch however much
data as could reasonably be expected from the crude
highlights with a specific end goal to segregate clusters of
phony records from clusters of honest to goodness
accounts.[4]The separated highlights can be
comprehensively gathered into three classes; which we
depict at abnormal state here; additionally subtle elements
can be found in Section.
Essential dispersion highlights. For each cluster; we take
essential factual measuresof every section (e.g.organization
name).[24]Cases incorporate mean or quartiles for
numerical highlights; or number of extraordinary esteems
for content highlights. 2. Example highlights. We have
outlined pattern en-coding calculations" that guide client
created content to a littler straight out space. We at that
point take essential dispersion includes over these straight
out factors.[22] These highlights are intended to identify
malevolent clients (particularly bots) that are following an
example in their air conditioning check information
exchanges. 3. Recurrence highlights.Foreachcomponent,we
register the recurrence of that incentive over the whole
record database. We at that point figure fundamental
dissemination includesover these frequencies.[10]Whenall
is said in done we expect clustersof genuine recordsto have
some high-recurrence informationandsomelow-recurrence
information, while bots or malignant clientswillindicateless
change in their information frequencies; e.g., utilizing just
normal or just uncommon names.[13]
1.3 Account Scorer
The Account Scorer's capacity is to prepare the models and
assess them on already inconspicuous information. The
Account Scorer takes information of Profile Featurizer; i.e.,
one numerical factor for each cluster is build.[20] The
extraordinary learning calculation utilized is client
configurable; in our investigations we think about strategic
relapse, arbitrary woods, and bolster vector machines. In
training mode," the Account Scorer is given a named set of
preparing information and yields a model portrayal and
assessment measurementsthat canbeutilizedtothinkabout
variousmodels.[9] In evaluation mode," the Ac-tallyScorer
is given a model depiction and an info vector of cluster
highlights and yields a score for that cluster showing the
probability of that cluster being made out of phony
accounts.[6] In light of the cluster's score, accounts in that
cluster can be chosen for any of three activities:
programmed confinement (if the likelihoodofbeingphonyis
high), manual survey (if the outcomes are uncertain), or no
activity (if the likelihood of being phony is low).[19] The
correct limits for choosing between the three activities are
designed to limit false positives and give human analysts a
blend of good and awful accounts. The physically named
accounts can later be utilized as preparing information in
advance emphases of the mode.[23]
2. BACKGROUND ANALYSIS
Techniques are devised to check the falsifying information
provided by the user in order to achieve desired goals.These
techniques are discussed in detail in this section.
2.1 Centralized Detection Technique
Grewal & Scholar 2015 Central authority is established in
detection and prevention of clone attack in social media.
Mainly this application is deployed in sensor networks but
due to heavy traffic associated with the socialmedianeedfor
centralized detection comes into place. Under centralized
detection following techniquesappear.[14]Khabbazianetal.
n.d.In this approach every node in the network send the
information towards the neighbor and neighbor in turns
sends the information towards the base station. The base
station scans all the relevant information from the received
report and if conflicting position information is spottedthen
message is conveyed to all the nodes in the network.[17]
Solanki 2016 Another method of detection is the set
operations. this mechanism reduces the number of
transmitted packets hence overheadisconsiderablyreduced
by the use of set operations. In this mechanism duplicity

from the subset is detected and removed. This technique is
cost effective but also disallows some of the critical packet
transmission.[25] Wang & Zhang 2007 Cluster based
approach is based on attribute similarity. The data being
transmitted is analysed for similarity based on properties
they have. In case of similar attributes disclosure, the
information is packed in a common group known as cluster.
Cluster contains information of similar sort hence these are
homogeneous clusters. Threshold value upon the numberof
attributes similarity is maintained. In case similarity index
exceeds threshold value, clone attack is detected.
Heterogeneous cluster is not worked upon in existing
literatures as yet.[29] Ren et al. n.d. This is the process in
which key generated through cryptography process is
analyzed for redundancy. The data transmitted in such
fashion is secured and difficult to analyze. In order to detect
replicated keys, extra storage is required atdatacenters.The
upcoming key is compared against the incoming keys. The
incoming keys if similar, replication is detectedsodoesclone
attack.[22]
2.2 Distributed Detection Techniques
These are the techniques which do not rely on the
centralized authority to check for the abnormality. Watcher
nosed are established in order to check for the anomalies.
Under distributed techniquesfollowing mechanismsappear
Devi & Poovammal 2016 Content based filtering mechanism
is used to detect the abnormal material among the
transmitted contents. The social media is prone to large
number of users having large numberofdataassociatedwith
them. Content filtering mechanism maintains a word count
register, containing the total number of words to be
transmitted. After transferringtheword,wordcountregister
is decremented by one. After word count register reaches 0,
words to be transmitted still pending is analyzed.Incaseany
word is still left, that indicates malicious entry. Some work
towards saving time is still required to be
accomplished.[11]Dave et al. n.d. Attributes are the
properties associated with the content being transmitted.
Attributes of use must be transmitted with integrity check.
Primary key is implied over the attributesbeingtransmitted.
The attributes values cannot be redundant andalsoitcannot
be null. Problem with the attribute based approach is
attribute similarity is checked but content is not purified.
[11]Mohammed et al. 2014 The attention on this paper is to
assemble an Android stage based portable application for
the medicinal services area, which utilizes the possibility of
Internet of Things (IoT) and distributed computing.Wehave
constructed an application called 'ECG Android App' which
gives the end client perception of their Electro Cardiogram
(ECG) waves and information logging usefulnessoutofsight.
The logged information can be transferred to the client's
private incorporated cloud or a particular restorative cloud,
which keeps a record of all the observed informationandcan
be recovered for investigation by the therapeutic staff.
Despite the fact that building a restorative application
utilizing IoT and cloud procedures isn't absolutely new,
there is an absence of observational investigations in
building such a framework. This paper surveys the major
ideasof IoT. Further, the paper displaysa foundation for the
medicinal services area, which comprises of different
advances: IOIO microcontroller, flag preparing,
correspondence conventions, secure and proficient
instruments for expansive record exchange, information
base administration framework, and the concentratedcloud.
The paper accentuateson the framework and programming
engineering and outline which is basic to general IoT and
cloud based restorative applications. The framework
exhibited in the paper can likewise be connected to other
medicinal services spaces. It finishes up with suggestions
and extensibilities found for the arrangement in the human
services space.[21] Tekieh & Raahemi 2015 In this review,
we gather the related data that show the significance of
information mining in social insurance. As the measure of
gathered wellbeing information is expanding fundamentally
consistently, it is trusted that a solid investigation device
that is equipped for taking care of and dissecting extensive
wellbeing information is basic. Breaking downthewellbeing
datasets assembled by electronic wellbeing record (EHR)
frameworks, protection claims, wellbeing studies, and
different sources, utilizing information mining methods is
exceptionally perplexing and is looked with certain
difficulties, including information qualityandsecurityissues.
In any case, the uses of information mining in social
insurance, focal points of information mining procedures
over conventional strategies, uncommon qualities of
wellbeing information, and new wellbeing conditionpuzzles
have made information digging extremely fundamental for
wellbeing information examination.[26] Kiruthiga 2014
Social Networks (SN) are prominent among the individuals
to associate with their companions through the web. Clients
investing their energy in prevalent social systems
administration destinations like facebook, Myspace and
twitter to share the individual data. Cloning assault is one of
the slippery assaults in facebook. Generally aggressors stole
the pictures and individual data about a man and make the
phony profile pages. Once the profile kicks cloned they offto
send a companion ask for utilizing the cloned profile. In the
event that if the genuine clients account gets blocked, they
used to send another companion demand to their
companions. In the meantime cloned one likewise sending
the demand to the individual. Around then it was difficult to
recognize the genuine one for clients. In the proposed
framework the clone assault is distinguished inlightofclient
activity era and clients click example to discover the
comparability between the cloned profileandgenuineonein
facebook. Utilizing Cosine closeness and Jaccard file the
execution of the similitude between the clients is made
strides.[18].
Both content based and attribute based approaches
commonly used with the applications of recommender
system.

3. COMPARISON OF VARIOUS TECHNQIUES FOR
CLONE ATTACK DETECTION
The comparison table for detection of clone attackisgivenas
under
Authors
and Year
Techniques Attack
Detected
Merit and Demerits
(Tsikerdekis
& Zeadally
2014)[28]
Nonverbal
Behavior
Multiple
Identities
Clone
attack
Detection
Non verbal behavior
techniques is implied
which gives result
faster but it may not be
accurate in all
situations
(Egele et al.
2015)[12]
Detection
using
similarity
profilecheck
Clone
Attack
Suited only for high
profile accounts while
low profile attacks are
difficult to identify
(Wu et al.
2017)[30]
Social Norm
Incentives
Sybil
attacks in
networks
Suitable for small
networks but is not
suited for complex
networks
(Anjos et al.
2014)[2]
Attack
detection
using face
recognition
Photo
Attack
detection
Used only for photo
attack in social media
(Amerini et
al. 2011)[1]
Copy move
attack
SIFT Based
mechanism
for attack
detection
Can be implied on large
image sets but not
tested on textual
information
(Shi et al.
2017)[24]
Event attack
detection
Event
detectionin
social
media
Fixed datasets or static
datasets uses produce
effective results but
dynamic datasets still
not checked
Table 1: Comparison of attack detection strategies
4. CONCLUSION AND FUTURE SCOPE
From the analysis conducted we conclude thatthedeception
is a common problem in the field of online social media. The
steps must be taken in order to prevent the deception. The
causes of deception is lack of structure to ensure that only
valid users can enter into the system. The proper validation
mechanisms are missing since the OSNistypicallyconcerned
about the length of the database rather than security of the
system. This is a prime factor which is leading to the
deception.
In the future some sort of security mechanisms must be
enforced to ensure the validity of the user. This can be
accomplished by the use of background check mechanisms
to prevent clone attacks.
REFERENCES
[1] Amerini, I. et al., 2011. A SIFT-Based Forensic Method
for Copy – Move Attack Detection and Transformation
Recovery. , 6(3), pp.1099–1110.
[2] Anjos, A., Chakka, M.M. & Marcel, S., 2014. Motion-based
counter-measuresto photo attacksin face recognition.
, (November 2012), pp.147–158.
[3] Aprem, A. & Krishnamurthy, V., 2016. Utility Change
Point Detection in Online Social Media : A Revealed
Preference Framework. , (c), pp.1–12.
[4] B, J.R. et al., 2016. Detecting Overlapping Community in
Social. , pp.99–110.
[5] Barbera, M. V & Mei, A., 2012. Personal Marks and
Community Certificates : Detecting Clones in Wireless
Mobile Social Networks.
[6] Bhat, S.Y. & Abulaish, M., 2014. Communities A gainst
Deception in Online Social Networks 1 The Platform 2
The Mischef. , 2014(2), pp.8–16.
[7] Bu, K. et al., 2015. Deterministic Detection of Cloning
Attacks for Anonymous RFID Systems. , 11(6),
pp.1255–1266.
[8] Caton, S. et al., 2014. A Social Compute Cloud : Allocating
and Sharing Infrastructure Resources via Social
Networks. , 1374(c), pp.1–14.
[9] Choi, S., Chung, K. & Yu, H., 2013. Fault tolerance andQoS
scheduling using CAN in mobile social cloud
computing.
[10] Dave, D., Mishra, N. & Sharma, S., Detection Techniques
of Clone Attack on Online Social Networks :Surveyand
Analysis. Elsevier, pp.179–186.
[11] Devi, J.C. & Poovammal, E., 2016. An Analysis of
Overlapping Community Detection Algorithms in
Social Networks. Procedia - Procedia Computer
Science, 89, pp.349–358. Available at:
http://dx.doi.org/10.1016/j.procs.2016.06.082.
[12] Egele, M. et al., 2015. TowardsDetecting Compromised
Accounts on Social Networks. , 5971(c).
[13] Etter, M. et al., 2017. Measuring Organizational
Legitimacy in Social Media : Assessing Citizens ’
Judgments With Sentiment Analysis.
[14] Grewal, R. & Scholar, P.G., 2015. A Survey on Proficient
Techniquesto Mitigate Clone AttackinWirelessSensor
Networks. , pp.1148–1152.
[15] Gupta, S. & Arora, S., A Hybrid Firefly Algorithm and
Social Spider Algorithm for Multimodal Function. ,
pp.17–30.
[16] Hovy, D. & Spruit, S.L., 2001. The Social Impact of
Natural Language Processing.

[17] Khabbazian, M., Mercier, H. & Bhargava, V.K.,
Wormhole Attack in Wireless Ad Hoc Networks :
Analysis and Countermeasure.
[18] Kiruthiga, S., 2014. Detecting Cloning Attack in Social
Networks Using Classification and Clustering
Techniques.
[19] Kontaxis, G. et al., Detecting Social Network Profile
Cloning.
[20] Maio, C. De et al., 2017. Unfolding social content
evolution along time and semantics.FutureGeneration
Computer Systems, 66, pp.146–159. Available at:
http://dx.doi.org/10.1016/j.future.2016.05.039.
[21] Mohammed, J. et al., 2014. Internet of Things: Remote
Patient Monitoring Using Web Services and Cloud
Computing. In 2014 IEEE International Conference on
Internet of Things(iThings), and IEEE Green
Computing and Communications(GreenCom)andIEEE
Cyber, Physical and Social Computing (CPSCom).IEEE,
pp. 256–263. Available at:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.ht
m?arnumber=7059670 [Accessed January 26, 2016].
[22] Ren, Y., Chen, Y. & Chuah, M.C., Social Closeness Based
Clone Attack Detection for Mobile Healthcare System.
[23] Rizi, F.S., Khayyambashi, M.R. & Kharaji, M.Y., 2014. A
New Approach for Finding Cloned Profiles in Online
Social Networks. , 6(April), pp.25–37.
[24] Shi, L. et al., 2017. Event Detection and User Interest
Discovering in Social Media Data Streams. , 3536(c).
[25] Solanki, S., 2016. Related Study of Soft Set and Its
Application A Review. , 7(4), pp.15–22.
[26] Tekieh, M.H. & Raahemi, B., 2015. Importance of Data
Mining in Healthcare. Proceedings of the 2015
IEEE/ACM International Conference on Advances in
Social Networks Analysis and Mining 2015 - ASONAM
’15, pp.1057–1062.
[27] Tsikerdekis, M. & Ieee, M., 2017. Real-Time Identity
Deception Detection Techniques for Social Media :
Optimizations and Challenges.
[28] Tsikerdekis, M. & Zeadally, S., 2014. Multiple Account
Identity Deception Detection in Social Media Using
Nonverbal Behavior. IEEETransactionsonInformation
Forensics and Security, 9(8), pp.1311–1321. Available
at:
http://ieeexplore.ieee.org/articleDetails.jsp?arnumber
=6843931 [Accessed February 25, 2016].
[29] Wang, W. & Zhang, Y., 2007. On fuzzy cluster validity
indices. Fuzzy Sets and Systems, 158(19), pp.2095–
2117.
[30] Wu, C., Gerla, M. & Schaar, M. Van Der, 2017. Social
Norm Incentives for Network Coding in MANETs. ,
pp.1–14.
[31] Zhou, H.W.J.L.L., 2011. Lightweight and effective
detection scheme for node clone attack in wireless
sensor networks. , (December 2010), pp.137–143.

IRJET- Competitive Analysis of Attacks on Social Media

More Related Content

What's hot (17)

Similar to IRJET- Competitive Analysis of Attacks on Social Media (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Competitive Analysis of Attacks on Social Media