Fast Binary Network Intrusion Detection based on Matched Filter Optimization

Securing networks has become very critical task because of the continued appearance of attacks and the growing number of Internet users. The detection, classification and prevention of attacks are provided by the so-called Intrusion Detection System (IDS). In this article, we have proposed and evaluated a new model of network intrusion detection based on matched filter optimization called NIDeMFO for Network Intrusion Detection based on Matched Filter Optimization. Similar to Linear Discriminant Analysis (LDA), the goal is to design a linear filter that projects data into a space where both classes, normal and attack, are well separated. The difference with LDA is that the margin between the averages of the two classes in the projected space is controlled by a parameter. The proposed detection model is evaluated on the NSL-KDD benchmark. The results show its competitiveness and effectiveness compared to many existing detection models.