A Distributed Intrusion Detection System Based on Bayesian Alarm Networks

Abstract

Intrusion Detection in large network must rely on use of many distributed agents instead to one large monolithic module. Agents should have some kind of artificial intelligence in order to cope successfully with different intrusion problems. In this paper, we suggested Bayesian alarm network to work as independent Network Intrusion Detection Agent. We have shown that when narrowed in detecting one specific type of the attack in large network, for example denial of service, virus, worm or privacy attack, we can induce much more prior knowledge into system regarding the attack. Different nodes of the network can develop their own model of Bayesian alarm network and agents could communicate between themselves and with common security data base. Networks should be organized hierarchically so on the higher level of hierarchy, Bayesian alarm network, thanks to interconnections with lower level networks and data, acts as a distributed Intrusion Detection System.