Correlation Power Analysis of Pipelined and Multi-Threaded Coarse-Grained Reconfigurable Cryptographic Accelerator

LiCryptor, our state-of-the-art coarse-grained reconfigurable cryptographic accelerator, is designed for emerging Internet of Things (IoT) systems that demand high-speed and energy-efficient hardware capable of supporting multiple cryptographic algorithms. In this paper, we evaluate the vulnerability of LiCryptor to Correlation Power Analysis (CPA) side-channel attacks (SCA), proposing two primary attack models based on Hamming Weight (HW) and Hamming Distance (HD) to target intermediate values stored in registers during encryption. We introduce 8 power model schemes for CPA attacks, covering both single-threaded pipelined operation and dual-threaded pipelined operation on LiCryptor. Experimental results show that, under the pipelined operation, the proposed HD-based energy model can reveal all 16 key bytes with 82,100 traces, achieving a $1.4 x$ improvement over single-threaded methods. In dual-threaded operation, the number of traces required for key recovery is reduced to 54,000 traces. Furthermore, we identify CPA_HD_R10 as the most effective attack model, outperforming other models in both single-threaded and dual-threaded configurations.