Danger theory metaphor in artificial immune system for system call data

Abstract

Artificial Immune System (AIS) is a naive paradigm in biologically inspired computation; artificial neural networks (ANNs) and genetic algorithms (GAs) are among popular examples in this domain. The field of AIS research is vast and complex that demands immense multi-disciplinary efforts. As AIS is designed on the principles of natural Immune System (IS); so features of immune-inspired computational metaphors reflect features of the immunological theories/phenomena upon which these metaphors are based. In immunology, there are two distinct viewpoints about main goal of IS; self-non-self and danger theory. Most of the existing AIS are based on classical self-non-self perspective. A recent recommendation has initiated some efforts exploring potentials of danger theory (DT) for AIS. A few existing DT based AIS metaphors are not sufficient to justify potentials of the vast field, so more explorations are needed. This study aims to contribute for the domain proposing a novel metaphor DASTON (DAnger Susceptible daTa codON). The effort completes four objectives; framework for abstracting immunology inspired computational metaphor, mechanism for DASTON abstraction, verifying existence of DASTON through benchmark data, and discovering novel biological property bio fitness for computational metaphors. Although, AIS is emerging as general paradigm for wide application area, computer security is its naturally analogous domain. So, exploitation of system call data, having enormous significance in computer security, is a good suggestion for this study. It concludes that; proposed framework is viable for abstracting immuneinspired metaphors, abstracted metaphor DASTON exists in system call data and fulfils proposed test criterion bio-fitness that proves its analogy to basis biological phenomena. The study also proposes a distinctive biological phenomenon danger susceptibility that might provide base for some useful immunological exploration. Hence, this thesis mainly contributes for DT based AIS with partial contributions for computer security, bio-inspired computation, and immunology