Real-time intrusion prevention and security analysis of networks using HMMs

In this paper we propose to use a hidden Markov model (HMM) to model sensors for an intrusion prevention system (IPS). Observations from different sensors are aggregated in the HMM and the intrusion frequency security metric is estimated. We use a Markov model that captures the interaction between the attacker and the network to model and predict the next step of an attacker. A new HMM is created and used for updating the estimated system state for each observation, based on the sensor trustworthiness and the time since last observation processed. Our objective is to calculate and maintain a state probability distribution that can be used for intrusion prediction and prevention. We show how our sensor model can be applied to an IPS architecture based on intrusion detection system (IDS) sensors, real-time traffic surveillance and online risk assessment. Our approach is illustrated by a small case study.