An Agent-Based Framework for Simulating and Analysing Attacks on Cyber Physical Systems

Abstract

An agent-based framework is presented to model and analyze physical and cyber attacks on Cyber Physical Systems (CPS). In the first phase of a two phase procedure in the framework, a CPS is modelled using State Condition Graphs (SCGs) representing the structural relations among the cyber and physical components. In the second phase, SCGs are partitioned into separate groups, referred to as agents, based on their functional dependency. Each agent is associated with a sensor and an actuator set. The actuator set is further divided into two subsets: active actuator and passive subsets that, respectively, directly or indirectly participate in controlling the environment. Interactions among different agents can be created by assigning an aggregate function to each agent with sensor and actuator sets as the domain of the function and status of the environment as its codomain. A case study was conducted on a segment of a 6-stage Secure Water Treatment (SWaT) system to demonstrate the applicability of the framework. The study reveals promising features of the framework to analyze the steady state response of the system following both cyber and physical attacks.