Vulnerability analysis of AIS-based intrusion detection systems using genetic and evolutionary hackers
Brown, Douglas
2005-05-01
2000-2009
In this thesis, an overview of current intrusion detection methods, evolutionary computation, and immunity-based intrusion detection systems (IDSs) is presented. An application named Genetic Interactive Teams for Intrusion Detection Design and Analysis (GENERTIA) is introduced which uses genetic algorithm (GA)-based hackers known as a red team in order to find vulnerabilities, or holes, in an artificial immune system (AlS)-based IDS. GENERTIA also uses a GA-based blue team in order to repair the holes it finds. The performance of the GA-based hackers is tested and measured according to the number of distinct holes that it finds. The GA-based red team�s behavior is then compared to that of 12 variations of the particle swarm optimization (PSO)-based red team named SWO, SW0+, SW1, SW2, SW3, SW4, CCSWO, CCSW0+, CCSW1, CCSW2, CCSW3, and CCSW4. Each variant of the PSO-based red team differs in terms of the way that it searches for holes in an IDS. Through this test, it is determined that none of the red teams based on PSO perform as well as the one based on a GA. However, two of the twelve PSO-based red teams, CCSW4 and SW0+, provide hole finding capabilities closest to that of the GA. In addition to the ability of the different red teams to find holes in an AlS-based IDS, the search behaviors of the GA-based hackers, PSO-based hackers that use a variable called a constriction coefficient, and PSO-based hackers that do not use the coefficient are compared. The results of this comparison show that it may be possible to implement a red team based on a hybrid �genetic swarm� that improves upon the performance of both the GA- and PSO-based red teams.
text�
application/pdf
thesis
Master of Science (MS)
Clark Atlanta University
Department of Computer Science
George, Roy
Georgia--Atlanta
http://hdl.handle.net/20.500.12322/cau.td:2005_brown_douglas
http://rightsstatements.org/vocab/InC-EDU/1.0/