Detection of traffic volume anomalies by evolution of negative classifiers in Artificial Immune Systems

Traffic volume anomalies can take a wide range of different forms, each characterized in principle by a different traffic profile, but all the forms having in common the overall surge in traffic at a particular site. Often anomalies, at the onset, appear up as innovations, an unprecedented experience for the network system. For this reason it is appropriate to face them with a negative selection approach that can detect foreign patterns in the complement space. In this work we propose to detect the onset of traffic anomalies within the paradigmatic approach of evolutionary artificial immune systems, through the use of classifiers evolved on the basis of normal traffic profile (the complementary space corresponds to the immune system non-self): the overall architecture can provide robustness and adaptability. The approach discussed here could apply not only to volume anomalies but to traffic anomalies in general.