AI-Assisted Electronic Digital Twin for Liquid Level Sensors under Hardware Trojan Attack: An Experimental Validation

This work proposes an Electronic Digital Twin (EDT) circuit model for Hardware Trojan (HT) attack characterization. Specifically, the goal is to compare circuit simulations, Artificial Intelligence (AI)-based attack classification and measurements proving a liquid level sensor attack by coupling an undesired resistance between the output nodes of the sensor. The paper describes all the steps to create the AI-assisted EDT model and validates the proposed modelling approach through experimental and numerical results. The paper adopts open-source tools, namely LTSPICE for circuit simulation, KNIME for AI classification, and Arduino IDE for firmware implementation. Concerning AI classification, the Random Forest (RF) supervised algorithm exhibited good performance for the classification of electrical resistance attacks. Theoretical and experimental behavior are investigated to validate the approach thus proposing a general protocol to realize an AI-assisted EDT framework.