Artificial Immune Systems for Self-Nonself Discrimination: Application to Anomaly Detection

Summary

Self-nonself discrimination is the ability of the vertebrate immune systems to distinguish between foreign objects and the body’s own self. It provides the basis for several biologically inspired approachs for classification. The negative selection algorithm, which is one way to implement self-nonself discrimination, is becoming increasingly popular for anomaly detection applications. Negative selection makes use of a set of detectors to detect anomalies in input data. This chapter describes two very successful negative selection algorithms, the self-organizing RNS algorithm and the V-detectors algorithm, which are useful with real valued data. It also proposes two new approaches, the single and the multistage proliferating V-detector algorithms to create such detectors. Comparisons with artificial fractal data as well as with real data pertaining to power distribution failure rates, shows that while the RNS and the V-detector algorithms can perform anomaly detection quite well, the proposed mechanism of proliferation entails a significant improvement over them, and can be very useful in anomaly detection tasks.