No abstract

Masquerade Detection Based On UNIX Commands by Amruta Mahajan In this paper, we consider the problem of masquerade detection based on a UNIX system. A masquerader is an intruder who tries to remain undetected by impersonating a legitimate user. Masquerade detection is a special case of the general intrusion detection problem. We have collected data from a large number of users. This data includes information on user commands and a variety of other aspects of user behavior that can be used to construct a profile of a given user. Hidden Markov models have been used to train user profiles, and the various attack strategies have been analyzed. The results are compared to a standard dataset that offers a more limited view of user behavior. ACKNOWLEDGMENTS I am grateful and take this opportunity to sincerely thank my thesis advisor, Dr. Mark Stamp, for his constant support, invaluable guidance, and encouragement. His work ethic and constant endeavor to achieve perfection have been a great source of inspiration. I wish to extend my sincere thanks to Dr. Sami Khuri and Dr. Chris Pollett for consenting to be on my defence committee and for providing invaluable suggestions to my project without which this project would not have been successful. I also would like to thank my husband, Neeraj Mahajan, for his support and encouragement throughout my graduation. v