AI-Based Detection and Simulation of Hardware Trojan Attacks in Industrial Applications

Industrial Internet of Things (IIoT) is becoming increasingly important for companies worldwide since it allows integrating data from different areas and sectors, thus creating new opportunities to raise efficiency, flexibility, and productivity. However, IIoT also exposes companies to new risks and threats, which implies new challenges related to cybersecurity. In this context, this paper analyzes Hardware Trojan (HT) attacks also characterized by possible malicious controlled white noises on sensors working in an IIoT scenario. Through a parametric analysis in a simulation environment, the impact and the sensitivity of the hardware attacks about the electrical signal deterioration is analyzed as a function of the variation of the HT resistance and capacitance, also superposing a white noise effect to increase the signal disturbance and complexity to find possible causes. The model also includes the possibility for the attackers to modulate and switch the white noise signal in accordance with the IIoT transmission triggering. Specifically, the paper analyzes the behavior of an attacked resistive humidity sensor by means of circuit simulations. Furthermore, the paper proposes a new approach for training commonly used Artificial Intelligence (AI) algorithms, such as Artificial Neural Networks (ANNs) and Random Forest (RF), to predict disturbed sensor signals. Simulation results show the proposed method is effective in detecting HT attacks, thus supporting the adoption of proper corrective actions. The AI-based data processing also demonstrates that circuit simulations provide an effective solution to train AI models tailored to a digital twin HT cybersecurity framework.