Conditional Random Fields for Intrusion Detection

An intrusion detection system is now an inevitable part of any computer network. With the ever increasing number and diverse type of attacks, including new and previously unseen attacks, the effectiveness of an intrusion detection system is often subjected to testing. The use of such systems have greatly reduced the threat level, however, the networks and hence the data and services offered by them are far away from the state when they can be considered as secure. In this paper we propose and experimentally validate the use and robustness of ´conditional random fields,´ for the task of intrusion detection. We show, experimentally, that the conditional random fields, can be very effective in detecting intrusions when compared with the previously known techniques.