Negative Selection with Antigen Feedback in Intrusion Detection

One of the major challenges for negative selection is to efficiently generate effective detectors. The experiment in the past shows that random generation fails to generate useful detectors within acceptable time duration. In this paper, we propose an antigen feedback mechanism for generating the detectors. For an unmatched antigen, we make a copy of the antigen and treat it the same as a newly randomly generated antibody: it goes through the same maturing process and is subject to elimination due to self matching. If it survives and is then activated by more antigens, it becomes a legitimate detector. Our experiment demonstrates that the antigen feedback mechanism provides an efficient way to generate enough effective detectors within a very short period of time. With the antigen feedback mechanism, we achieved 95.21% detection rate on attack strings, with 4.79% false negative rate, and 99.21% detection rate on normal strings, 0.79% false positive. In this paper, we also introduce Arisytis – Artificial Immune System Tool Kits – a project we are undertaking for not only our own experiment, but also the research communities in the same area to avoid the waste on repeatedly developing similar software. Arisytis is available on the public domain. Finally, we also discuss the effectiveness of the r-continuous bits match and its impact on data presentation.