Modified PSO based Intrusion Detection in Streaming Network Data

Abstract: Intrusion Detection is one of the prime functionalities of any Network Based System and it reflects the quality of the network. The existing intrusion detection systems performs in a fairly predictable and reliable manner, however, on the other side, hackers and attackers are always ahead, and they keep conjuring new attacks in a constant and consistent manner. The only way to solve this issue is to detect or predict an attack as soon as possible, in real time and if possible to prevent them from happening all together. Detecting it after it has occurred is useless, as the damage is already done. However, the current systems used for detecting intrusions are not capable of providing real time protection against such attacks. Latencies in the detection structure has become unavoidable due to the hugeness of the data associated with it. This paper presents an Intrusion Detection System based on PSO, a well-known metaheuristic algorithm. Regular PSO is hybridized with Simulated Annealing to provide a more effective and faster detection mechanism. Experiments conducted on PSO-SA prove that the algorithm works effectively on networked data providing high quality results. Further, comparisons with regular PSO also indicate that the proposed technique performs much better than regular PSO in detecting intrusions.

Keywords: PSO; Simulated Annealing; Network Intrusion Detection; KDD.