A Study of Learning Method for Intrusion Detection System using Machine Learning

摘要

The network intrusion is becoming a big threat. Recent intrusions are becoming more clever and difficult to detect. Many of today's intrusion detection systems are signature-based. They have good performance for known attacks, but theoretically they are not able to detect unknown attacks. On the other hand, an anomaly detection system can detect unknown attacks and is getting focus recently. In this paper, we study the effectiveness and the performance experiments of one of the major anomaly detection scales, L0F, on distributed online machine learning framework, Jubatus. After basic experiment, we propose a new machine learning method and show our new method has better performance than the original method.