Design of microarchitectural security counters for detecting software and hardware compromises in critical embedded systems
Lucas Georget PhD defense
23.11.26 - 23.11.26
This thesis proposes a new security approach for industrial systems in the face of detecting insidious software and hardware compromises. While the state of the art classically exploits hardware performance counters for this purpose, the evolution of industrial systems towards generic architectures deployed on FPGAs enables the envisaging of fine-grain analysis methods, by observing internal signals that effectively discriminate malicious behaviors. This thesis proposes an end-to-end approach enabling the automated selection of internal system signals, dimensionality reduction via explainability methods (SHAP values), and lightweight integration through the implementation of decision trees. Two complementary case studies concerning software and hardware Trojans are presented to illustrate the genericity of the approach.
published on 17.11.26