A Machine Learning Based Framework for Real-Time Detection and Mitigation of Sensor False Data Injection Cyber-Physical Attacks in Industrial Control Systems

In light of the advancement of the technologies used in industrial control systems, securing their operation has become crucial, primarily since their activity is consistently associated with integral elements related to the environment, the safety and health of people, the economy, and many others. This work presents a distributed, machine learning based attack detection and mitigation framework for sensor false data injection cyber-physical attacks in industrial control systems. It is developed using the system's standard operational data and validated using a hybrid testbed of a reverse osmosis plant. A MATLAB/Simulink-based simulation model of the process validated with actual data from a local plant is used. The control system is implemented using Siemens S7-1200 programmable logic controllers with 200SP Distributed Input/Output modules. The proposed solution can be adopted in the existing industrial control systems and demonstrated effective performance in real-time detection and mitigation of actual cyber-physical attacks launched by compromising the communication links between the process and the programmable logic controllers.

Automated summary

This work presents a distributed, machine learning based attack detection and mitigation framework for sensor false data injection cyber-physical attacks in industrial control systems. The framework is developed using the system's standard operational data and validated using a hybrid testbed of a reverse osmosis plant. The proposed solution can be adopted in the existing industrial control systems and demonstrated effective performance in real-time detection and mitigation of actual cyber-physical attacks.