IRJET- Web Application Firewall: Artificial Intelligence ARC

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 04 | Apr 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 3706
Web Application Firewall: Artificial Intelligence Arc
Parikshit Prabhudesai1, Aniket A. Bhalerao2, Rahul Prabhudesai3
1Director, Pitambari Products Pvt. Ltd., Maharashtra, India
2Deputy General Manager, IT & System Department, Pitambari Products Pvt. Ltd., Maharashtra, India
3Assistant General Manager, IT & System Department, Pitambari Products Pvt. Ltd., Maharashtra, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract – Nowadays every business and individual, are
using online platform to do business and to promote
themselves by performing financial transactions as well as
handling user confidential information transactions. Every
website holder needs security against all knownandunknown
threats; hence, we are developing a web application firewall
using artificial intelligence architecture to recognize attacks
and existing vulnerabilities by experiencing the behavior of
attacker and user in a unique way.
1. INTRODUCTION
Web application firewall is necessary for all static and
dynamic website holders to maintain & enhance security of
information, which is available on website or on server. We
all know that attackers are finding various vulnerabilities
daily. We need to update our security system by giving it
self-intelligence by changing our approach towards
protection by applying self-created knowledgebase.
1.1 A. I. Architecture Engine
Artificial intelligence architecture engine has a base and
builds on the integrity parameters defined by OWASP2 &
ITProPortol3. It has automated prevention and mitigation
system which isabletorecognizeattackpatternbehaviorand
impact on the information system to identify attack pattern
by building own knowledgebase and mitigation category. Its
algorithm has the capability to skip false positive attack
pattern by building own testing environment lab for all new
packets to the software. The software identifies whether it’s
a false positive or a positive impact.
Artificial Intelligence architecture engine has OWASP2
integration module which will help to build predefined
vulnerability database as well as help to build a
knowledgebase for particular attack type.
1.2 OWASP2 Integration
OWASP2 (open web applicationsecurityprojects)isa web
portal, which keeps track on all vulnerabilities from the
globe and categorises it by its severity and impact on
information system, hence it is the leading open web
vulnerability database. OWASP has developed an API which
is able to provide data access for third party queries and to
get predefined database. In order to stay relevant with the
time, we have integrated OWASP API, so that wecanprovide
cutting edge security.
Table -1: OWASP TOP VULNERABILITIES CHART
Vulnerability Severity
Type
DDoS High Web Threat
Spamming
Medium Mail Threat
SQL Injection
High
Database Threat
Proxy
High
Identity Threat
2. Detection Method
2.1 Modules:
Detection method contains two uniquely designed modules
for threat detection and mitigation. In first module, when a
WAN packet approaches theDNS, it is automaticallydiverted
to the WAF. WAF then separates its segments depending
upon meta-data. The AI enginewillcheckthesourcecodeand
threat segments depending upon the defined database and
behavioral based AI engine knowledgebase. In second
module, mitigation is applied by using OWASP engine or by
using own created mitigation algorithm to treat packets
properly by removing false positive.
2.2 Algorithm:
In first phase, all packets are filtered through the main web
application firewall engine, which is integrated with OWASP
for filtering predefined vulnerabilities as well as identifying
and eliminating globally defined threats.
If OWASP definition matches to the input packets, then the
packet will be dropped immediately by WAF. If input packet
definition does not match with OWASP definition, then first
phase will mark those packets partially cleaned. Here, the
first phase will end and the packet will be transferred to the
second phase.
In second phase, AIenginewillreceivethosepartiallycleaned
packets as an input. First activity from AI engine will be to
record packets’ behavioral pattern and if behavioral pattern
matches to the existing knowledgebase then AI engine will
mark those packets as malicious and in another case, if
packets’ behavioral pattern does not match with existing
knowledgebase then AI engine will inspect the packets’
behavior by giving it virtual environmenttodetectwhetherit

is harmful for website or not. If packets behavioral pattern is
found harmful then AI engine will record packets behavioral
pattern and insert it to knowledgebase and if the packets
behavioral pattern is not found harmful to the virtual web
application then the AI engine will mark it to cleaned and
pass those cleaned packets tothird phase, here secondphase
ends.
In third phase WAF targets IP section and to apply global
filter on the IP’s, IP sanitization section includes IP ban
system which is regularly updated by taking updates from
Virustotal6 API. We have integrated IP sanitization section
with virustotal API which fortifies the database by giving the
information about the globally banned IPs to the system.
Finally, cleaned packet with cleaned IP will go to the main
web application and get response from web application but
before getting a response, each transaction of packets will be
recorded in a log retention system.
Chart -1: Flow Chart
2.3 Affected Area:
Fig -2: WAF Working
Basic working of the web application firewall is to prevent
malicious packets from reachingtothemainwebapplication.
It is not concerned about the vulnerabilities of the source
code and hosting vulnerabilities as this WAF has its own
independent detection system. The firewall will not be
affected even in case of multiple external vulnerabilities.
3. ADVANTAGES
1. Intrusion prevention system and intrusion
detection system will get its own artificial
intelligence as a backup layer, whichwill providean
advance layer to the threat protection system by
giving transaction wise experience to the system,
which will handle threats more carefully than ever
before to eliminate false positive results.
2. While analyzing and preventing threat or malicious
packets, normal firewall needs to scan each packet
separately and it takes more time comparatively.
By providing artificial intelligence to web
application firewall,whichmakessegmentscanning
on each packet, it takes very less time to handle
threat as well as very low bandwidth consumption.
3. Artificial intelligence will reduce manpower and
human interaction as well as human error by giving
experience to the system to handle each threat and
to reduce false positive response.
4. Threat log retention helps by maintaining
knowledgebase and taking actions actively by
learning from the knowledgebase.
4. APPLICATION
By doing existing market survey, we found that among all
CMS’s major parts are using PHP language as a web
application platform. So, based on this information we
implemented above algorithm into user friendly application
by using PHP language, which is open source. According to
this study, while developing userendapplication,wecreated
a process flow in three phases. In first phase, we create
neural network at the application layer for routing packets
through threat detection engine by making independent

proxy server for a particular layer after which segmentation
is applied on the packets to split into layers.
In second phase, we send those segmented packets to
artificial intelligence arc engine to detect behavioral pattern
from inbuilt knowledgebase and to save these behavioral
events in event log section and to follow algorithm so on.
In third phase, we integrate global threat detection system
with the main firewall engine to detect and drop globally
declared threat definitions; after which it passesclearedand
clean packets to IP sanitizationmoduletodetect whetherthe
IP is banned. If the IP is not banned it reaches to the web
application.
Fig -3: Application layer attack tool analysis
Above figure shows the statisticsofattackingmethods which
has an impact on the application layer by using various
destructive hacking tools. Therefore, our main target is to
protect the application layer.
5. CONCLUSION
Hence, we conclude that among all existing web application
firewalls, knowledgebase system with artificial intelligence
is not implemented yet. It is imperative that the, updated
security system must be armed with AI to recognize attack
pattern and behavior by creating own knowledgebase and
mitigation system by eliminating false positive results
separately and by treating each packet independently. We
have developed a neural network based AI engine for web
application firewall which is able to mitigate all the
loopholes by using artificial intelligence.
REFERENCES
[1] Web Application Firewall Market Worth$5.48Billionby
2022. CISO Magazine. 5 October 2017. Retrieved 10
April 2018.
[2] "Web ParameterTampering -OWASP".www.owasp.org.
[3] Svartman, Daniel (12 March 2018). "The OWASP Top
Ten and Today's Threat Landscape". ITProPortol.
Retrieved 10 April 2018.
[4] K. Elissa, “Title of paper if known,” unpublished. Jason
Pubal (March 13, 2015). "Web Application Firewalls -
Enterprise Techniques" (PDF). SANS Institute. SANS
Institute InfoSec Reading Room.
[5] "TEST METHODOLOGY Web Application Firewall 6.2".
NSS Labs. NSS Labs. Retrieved 2018-05-03.
[6] Lardinois, Frederic. "Google Acquires Online Virus,
Malware and URL Scanner VirusTotal". TechCrunch.
Retrieved 12 April 2013.
[7] Anderson, James P., "Computer Security Threat
Monitoring and Surveillance," Washing, PA, James P.
Anderson Co., 1980.
[8] David M. Chess; Steve R.White(2000)."AnUndetectable
Computer Virus". Proceedings of Virus Bulletin
Conference. CiteSeerX 10.1.1.25.1508.
[9] Denning, Dorothy E., "An Intrusion Detection Model,"
Proceedings of the SeventhIEEESymposiumonSecurity
and Privacy, May 1986, pages 119–131
[10] Lunt, Teresa F., "IDES: An Intelligent System for
Detecting Intruders," Proceedings of the Symposium on
Computer Security; Threats, and Countermeasures;
Rome, Italy, November 22–23, 1990, pages 110–121.
[11] "Comparison operators". PHP.net.
[12] Pawel Krawczyk (2013). "Most common attacks on web
applications". IPSec.pl. Retrieved 2015-04-15.
[13] Pawel Krawczyk (2013). "So what arethe"mostcritical"
application flaws? On new OWASP Top 10". IPSec.pl.
Retrieved 2015-04-15.
ABBREVIATIONS
[1] OWASP – Open Web Application Security Projects
[2] WAF – Web Application Firewall
[3] WAN – Wide Area Network (i.e. Internet)
[4] DNS – Domain Name Server
[5] AI – Artificial Intelligence
[6] IP – Internet Protocol
[7] API – Application Program Interface
[8] CMS – Content Management System
[9] PHP – Personal Home Page

IRJET- Web Application Firewall: Artificial Intelligence ARC

More Related Content

What's hot (20)

Similar to IRJET- Web Application Firewall: Artificial Intelligence ARC (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Web Application Firewall: Artificial Intelligence ARC