INTRUSION DETECTION USING GRAPHICAL PROBABILITIC MODEL

Abstract

Detecting intrusion in network and application has become one of the most critical tasks to prevent their misuse by attackers. The Probabilistic graphical model for intrusion detection is presented; the objective of this model is to develop an intrusion detection system which will significantly detect the intrusion in a network and application. With the ever increasing number and diverse types of attack, including new and previously unseen attacks, their is a need to develop a system which will detect all the four classes of attacks; denial of service attack, probe attack, user to root attack and remote to layer attack. A technique called linear chain conditional random field is used which is a framework for building a probabilistic graphical model for intrusion detection, this linear chain conditional random field may be viewed as an undirected graphical model, which offers several advantages over other probabilistic model like Hidden Markov model, Naïve bayes model and Maximum entropy model. Using this technique we have developed a probabilistic model based application and detected all the four classes of attacks using the domain knowledge of intrusion and the features of all the four classes of attacks.