Leveraging EM Side-Channels for Recognizing Components on a Motherboard

This paper proposes leveraging EM side-channels to recognize/authenticate electronic components integrated onto a motherboard. By focusing on components on a motherboard, our method provides an opportunity to authenticate devices assembled by third parties. This method identifies components based on the modulated signals emanated while they are excited in a controlled manner. When testing an unknown component, the spectrum is compared to previously recorded training signatures. To improve efficiency, the size of the spectrum is reduced by projecting it into a vector space generated from training signatures. The identity of the tested component is then determined using a k-Nearest Neighbors algorithm. This method successfully classified memory, processor, and Ethernet transceiver components integrated on seven types of Internet-of-Things devices. Since manufacturers commonly use the same components in multiple designs, cross-type testing of motherboards is conducted. Collecting the training signatures on one motherboard and testing components from different motherboards speeds up the process and decreases the cost. Using measurements taken while exciting the components for 1 s, our method achieves a classification accuracy greater than 96% across all components tested. These results demonstrate that this method can recognize components based on their emanations, even if the components are integrated onto completely different motherboards.