Multi-shaped Detector Generation Using Real Valued Representtion for Anomaly Detection

In presenting this thesis in partial fulfillment of the requirements for a Master's degree at The University of Memphis, I agree that the Library shall make it available to borrowers under rules of the Library. Brief quotations from this thesis are allowable without special permission, provided that accurate acknowledgement of the source is made. Permission for extensive quotation from or reproduction of this thesis may be granted by my major professor, or in [his/her] absence, by the Head of Interlibrary Services when, in the opinion of either, the proposed use of the material is for scholarly purposes. Any copying or use of the material in this thesis for financial gain shall not be allowed without my written permission. Acknowledgements To Dr. Dipankar Dasgupta, major professor and supervisor of this research for the wonderful opportunity given to me, for his continuous guidance and tireless support. for their valuable comments, helpful suggestions, and constructive criticism. Colombia, Colombia for his brilliant ideas and continued interest in this work that helped me clear several milestones and ultimately deliver my mission. for all their help and friendship that made my entire time much more enjoyable. To my buddies and family members for their unconditional love, support and encouragement in my endeavors through this process. Artificial Immune Systems (AIS) is a new paradigm of soft computing which is motivated by the Biological Immune System (BIS). Negative selection algorithm is one of the important techniques in this paradigm that is widely applied to solve two-class (self and non-self) classification problems. Many advances to Negative Selection Algorithms (NSA) occurred over the last decade. This algorithm uses only one class (self) for training resulting in the production of detectors for the complement class (non-self). This paradigm is very useful for anomaly detection problems in which only one class is available for training, such as intrusive network traffic and its detection problem. This work makes a review of the basic AIS models and the related representation schemes. Besides binary-valued Negative Selection algorithms studied in the past, the Real Valued Negative Selection algorithm is the focus on solving Anomaly detection problems in the recent works [29, 42, 57]. Some of the well known work that use this scheme are, Negative Selection with Detection Rules (Evolutionary algorithm with hyper-rectangle detectors), Real-valued Negative Selection (heuristic algorithm with hyper-spherical detectors), Randomized Real-valued Negative Selection (Randomized Monte-Carlo method with hyper-spherical detectors), V-Detector (Boundary-aware, non-evolutionary statistical …