Abstract :
Inspired by biological immune systems, a novel immunity-based model, referred to as IMAD, is presented. In the model, the formal definitions of self, non-self, detectors, immune tolerance, and etc., are given. Furthermore, the quantitative description of the detector diversity is introduced to improve the generating efficiency of memory detectors, to reduce the number of memory detectors, to enlarge the coverage of non-self space, and immune response and immune detection are described in IMAD. To determine the performance of IMAD, the experiments comparing with different anomaly detection methods, including Negative Selection Algorithm: NSM, multilevel immune learning algorithm: MILA and Variable-Sized Detectors Algorithm: V-detector, were performed. Results exhibited that IMAD outperforms the previous techniques.
Keywords :
artificial immune systems; security of data; anomaly detection; biological immune systems; detector diversity quantitative description; immunity-based model; memory detectors; Artificial immune systems; Biological information theory; Biological system modeling; Biology; Computer science; Detectors; Immune system; Machine learning algorithms; Probability; Software engineering;