Sensing danger: Innate immunology for intrusion detection

The immune system provides an ideal metaphor for anomaly detection in general and computer securities in particular. Based on this idea, artificial immune systems have been used for a number of years for intrusion detection, unfortunately so far with little success. However, these previous systems were largely based on immunological theory from the 1970s and 1980s and over the last decade our understanding of immunological processes has vastly improved. In this paper we present two new immune-inspired algorithms based on the latest immunological discoveries, such as the behaviour of Dendritic Cells. The resultant algorithms are applied to real-world intrusion problems and show encouraging results. Overall, we believe that there is a bright future for these next-generation artificial immune algorithms.