A Multi-Mutation Pattern Immune Network for Intrusion Detection

Basing on the immune network theory and pattern recognition approach, a multi-mutation pattern immune network (MPIN) adaptive detector is proposed. By utilizing the immune response principle, the detection algorithm is designed. Because new features can be learnt by the MPIN in the real-time way, the detector is able to modify dynamically without periodical updating, and the detector´s ability of identifying novel attacks are also improved. Combined with a template-adjustable decision templates fusion algorithm, a three-level-module adaptive intrusion detection system (TAIDS) is presented. Experiments are carried out on Fisher Iris dataset and KDD-CUP-99 database to verify the performance of this MPIN detector and TAIDS. Compared with the detection approach based on neural networks, the false positive rate is decreased by 17.43% and the detection accuracy of unknown attacks is increased by 24.27%.