Identifying Vulnerabilities of Industrial Control Systems using Evolutionary Multiobjective Optimisation

TL;DR

This paper introduces a novel method using evolutionary multiobjective optimization algorithms to identify vulnerabilities in industrial control systems, evaluate attack impacts, and develop defense mechanisms, demonstrated on a chemical plant simulator.

Contribution

The paper presents a new approach combining EMO algorithms with vulnerability analysis and defense testing for complex industrial control systems.

Findings

EMO algorithms effectively identify system vulnerabilities.

Generated combinatorial attacks cause system damage and economic loss.

The approach outperforms random attack strategies.

Abstract

In this paper we propose a novel methodology to assist in identifying vulnerabilities in a real-world complex heterogeneous industrial control systems (ICS) using two evolutionary multiobjective optimisation (EMO) algorithms, NSGA-II and SPEA2. Our approach is evaluated on a well known benchmark chemical plant simulator, the Tennessee Eastman (TE) process model. We identified vulnerabilities in individual components of the TE model and then made use of these to generate combinatorial attacks to damage the safety of the system, and to cause economic loss. Results were compared against random attacks, and the performance of the EMO algorithms were evaluated using hypervolume, spread and inverted generational distance (IGD) metrics. A defence against these attacks in the form of a novel intrusion detection system was developed, using a number of machine learning algorithms. Designed approach was further tested against the developed detection methods. Results demonstrate that EMO algorithms are a promising tool in the identification of the most vulnerable components of ICS, and weaknesses of any existing detection systems in place to protect the system. The proposed approach can be used by control and security engineers to design security aware control, and test the effectiveness of security mechanisms, both during design, and later during system operation.