Side-channel Information Leakage Analysis on Functional FPGA Intellectual Properties by Neural Network

The utilization of FPGA continues to rise. Studies have suggested that digital circuits may have the risk EMI side channel leakage concerns. However, there are relatively few discussions on the concerns targeted to FPGA containing different soft-cores. This paper verifies whether EMI leakage of IPs can be distinguished when they are programmed into an FPGA. This paper uses a customized measurement platform to perform EMI analyses of different FPGA soft-core architectures. These test architectures in this paper are CPU, IO control, CLOCK, counter, and FIFO. By classifying the EMI leakages through a neural network (NN) to infer its corresponding architecture, the NN can distinguish the five kinds of EMI after NN training. It is also confirmed that FPGA does have EMI side channel leakage concerns.