Detecting Network Attacks via Improved Iterative Scaling

Network security has become a critical issue with the rapid increase in connectivity of computer systems over the Internet which has resulted in a great deal of opportunities for intrusions. One commonly used defense measure against such malicious attacks in the Internet is Intrusion Detection System (IDS). In this paper we describe a new data mining based method for intrusion detection based on network connection features. This method attempts to separate different kinds of intrusions from normal activities by using Improved Iterative Scaling (IIS). In addition, we describe a Chi-squared based method for selecting relevant connection features to improve the performance. Experiments validating the feasibility of the approach are presented.