Automatic security assessment of control systems for critical cyber-infrastructures
Roy H. Campbell, Zahid Anwar · 2008
Government and advisory agencies have issued various critical infrastructure protection standards and best-practice schemes for security-hardening of power and control networks in the wake of several security incidents threatening the reliability of the Power Grid. These security schemes differ in their installation costs and the degree of protection they provide against cyber attacks. Manually determining vulnerabilities and security risks in control networks and their maintenance procedures is a cumbersome process. Furthermore the process of mapping security schemes to vulnerable cyber assets each with a unique criticality in the power network configuration can easily become intractable. This research attempts to bridge this gap by investigating the automation of security assessment of the static and dynamic properties of critical infrastructures. We describe first-order logic based models of the static elements including power and control devices, services and connectivity, and re-writing logic based models of the dynamic elements such as operating procedure workflows, and the state of a working power grid. We introduce a tool-chain that, with a little manual assistance, can automatically generate these models from specifications, continuously update attributes from online event aggregators, and perform security assessment. Aside from checking whether the system configuration conforms to recommended best-practices for establishing security controls, the assessment also reveals whether the observed anomalies about the system could indicate possible security problems and permits dynamic ranking of alternative recovery procedures to minimize the total risk. Moreover the tool-chain can recommend an optimal selection of security schemes to apply to various vulnerable parts of the Power Grid network to maximize security when faced with a budget constraint. A case study on security hardening the IEEE power system 118-bus test case from a pool of five different best-practice schemes is used to demonstrate the feasibility of the tool chain implementation.