IEEE Design & Test (D&T)
Advances in chip manufacturing technologies have enabled computer architects to utilize System-on-Chip (SoC) to integrate the intellectual property cores as well as other components. Network-on-Chip (NoC) is widely used to fulfill communication requirements in SoC architectures. Securing NoC is vital for designing trustworthy SoCs. Eavesdropping attacks can exploit NoC vulnerabilities to extract secret information. In this paper, we propose a machine learning based detection of eavesdropping attacks. Our machine learning models are trained offline and have been used for runtime detection with a collective decision making strategy. Experimental results demonstrate that our approach can provide high accuracy with minimal overhead.
