Systematical Evasion From Learning-Based Microarchitectural Attack Detection Tools
Published in IEEE Journal on Emerging and Selected Topics in Circuits and Systems (IEEE JETCAS), 2024
Machine learning-based detectors have been proposed as a promising defense against microarchitectural attacks by classifying hardware performance counter (HPC) traces. In this paper, we systematically study the robustness of such detectors and present a general evasion framework that uses adversarial perturbations to modify attack code behavior while maintaining its effectiveness. Our analysis covers multiple detector architectures and attack types, revealing fundamental limitations in the security of learning-based detection approaches.
Recommended citation: D. R. Dipta, J. Tan and B. Gulmezoglu, "Systematical Evasion From Learning-Based Microarchitectural Attack Detection Tools," IEEE Journal on Emerging and Selected Topics in Circuits and Systems (IEEE JETCAS), vol. 14, no. 4, pp. 823-833, Dec. 2024.
Download Paper