Using Hidden Markov Models to Evaluate the Risks of Intrusions System Architecture and Model Validation

摘要

Security-oriented risk assessment tools are used to determine the impact of certain events on the security status of a network. Most existing approaches are generally limited to manual risk evaluations that are not suitable for real-time use. In this paper, we introduce an approach to network risk assessment that is novel in a number of ways. First of all, the risk level of a network is determined as the composition of the risks of individual hosts, providing a more precise, fine-grained model. Second, we use Hidden Markov models to represent the likelihood of transitions between security states. Third, we tightly integrate our risk assessment tool with an existing framework for distributed, large-scale intrusion detection, and we apply the results of the risk assessment to prioritize the alerts produced by the intrusion detection sensors. We also evaluate our approach on both simulated and real-world data.