Real-Time Intrusion Prediction Using Hidden Markov Model with Genetic Algorithm

As the use of Internet increases, cyber attacks and their severity also increase. Since it is not possible to compromise on security, intrusion detection systems (IDSs) become critical component in a secure organization. IDSs detect an attack only after it has occurred. When use in a high-traffic network, IDSs produce a large number of alerts. The false-positive (FP) rate increases with this. In this paper, we propose a framework for predicting future attacks by combining two machine-learning methods: genetic algorithm (GA) and hidden Markov model (HMM). It has two major components in which the first component makes use of GA to derive efficient intrusion detection rules and thereafter a precise detection of attacks. The second component uses HMM to predict the next attack class of the attacker. So combining these together is a good idea and gives a good intrusion prediction capability with reduced FP rate.