Integrating IT and OT for Cybersecurity: A Stochastic Optimization Approach via Attack Graphs

Abstract

This paper proposes an attack graph-based optimization model to enable cybersecure digital manufacturing. Cybersecurity has become imperative as manufacturing systems continue to increase connectivity through Industrial Internet of Things (IIoT) devices. However, modeling cyber threats in manufacturing environments remains underexplored. This work addresses this gap by presenting an approach to represent a manufacturing IT and OT network as an attack graph that captures vulnerabilities in components, such as the motion control system, spindle, tool changer, sensors, network interfaces, and connectivity through potential vectors. A two-stage stochastic programming model is formulated based on the attack graph to optimize the allocation of countermeasures under budget constraints considering distinct defense strategies to minimize expected cyber risk. A hybrid solution approach that integrates the Sample Average Approximation (SAA) and Benders Decomposition (BD) algorithms is used to efficiently solve the resulting large-scale problem. We demonstrate the efficacy of this proposed approach in cybersecure digital manufacturing via a real-life Computer Numerical Control (CNC) machining process. The computational results demonstrate that the proposed solution approach can solve the problem for a large-scale complex network, compromising more than 1000 connections between components, within 13 minutes. The defense strategies identified by our approach demonstrate that robust security protection can be achieved with optimal resource allocation, providing robust protection while minimizing implementation costs across the most critical vulnerabilities in the manufacturing network.