Electronic Theses and Dissertations

Identifier

1164

Date

2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Computer Science

Committee Chair

Sajjan G. Shiva

Committee Member

Dipankar Dasgupta

Committee Member

Vasille Rus

Committee Member

Linda Sherrell

Abstract

Cyber attacks have greatly increased over the years and the attackers have progressively improved in devising attacks against specific targets. Cyber attacks are considered a malicious activity launched against networks to gain unauthorized access causing modification, destruction, or even deletion of data. This dissertation highlights the need to assist defenders with identifying and defending against cyber attacks. In this dissertation an attack issue resolution system is developed called AVOIDIT IRS (AIRS). AVOIDIT IRS is based on the attack taxonomy AVOIDIT (Attack Vector, Operational Impact, Defense, Information Impact, and Target). Attacks are collected by AIRS and classified into their respective category using AVOIDIT.Accordingly, an organizational cyber attack ontology was developed using feedback from security professionals to improve the communication and reusability amongst cyber security stakeholders. AIRS is developed as a semi-autonomous application that extracts unstructured external and internal attack data to classify attacks in sequential form. In doing so, we designed and implemented a frequent pattern and sequential classification algorithm associated with the five classifications in AVOIDIT. The issue resolution approach uses inference to educate the defender on the plausible cyber attacks. The AIRS can work in conjunction with an intrusion detection system (IDS) to provide a heuristic to cyber security breaches within an organization. AVOIDIT provides a framework for classifying appropriate attack information, which is fundamental in devising defense strategies against such cyber attacks. The AIRS is further used as a knowledge base in a game inspired defense architecture to promote game model selection upon attack identification. Future work will incorporate honeypot attack information to improve attack identification, classification, and defense propagation.In this dissertation, 1,025 common vulnerabilities and exposures (CVEs) and over 5,000 lines of log files instances were captured in the AIRS for analysis. Security experts were consulted to create rules to extract pertinent information and algorithms to correlate identified data for notification. The AIRS was developed using the Codeigniter [74] framework to provide a seamless visualization tool for data mining regarding potential cyber attacks relative to web applications. Testing of the AVOIDIT IRS revealed a recall of 88%, precision of 93%, and a 66% correlation metric.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS