Improvement of Protocol Anomaly Detection Based on Markov Chain and Its Application

摘要

As we know, a lot of network attacks come from abusing different network protocols and several new attacks violate the protocol standard. Kumar Das first presented the concept of the protocol anomaly detection. The idea of protocol anomaly detection is not new but interesting. It aims to set up models for proper use of protocols and any behavior that departs from the models will be regarded as an intrusive or suspicious one. In this paper, we made some improvements that aim at the lack of stochastic protocol models based on Markov Chain and made some evaluations for that presented by Juan M. Some necessary states are added to the protocol model. Furthermore, the initial and transition probabilities are more precise. Also, we propose to combine Chi-Square Distance into Markov Chain method to detect protocol anomaly. The experimental results show that SYN Flooding attack can be detected efficiently by the new approach.