Hardware Acceleration for Cryptography Algorithms by Hotspot Detection

Data Encryption/Decryption has become an essential part of pervasive computing systems. However, executing these cryptographic algorithms often introduces a high overhead. In this paper, we select nine widely used cryptographic algorithms to improve their performance by providing hardware-assisted solutions. For each algorithm, we identify the software performance bottleneck, i.e., those “hotspot functions” or “hot-blocks” which consume a substantial portion of the overall execution time. Then, based on the percentage of execution time of a specific function and its relationship with the overall algorithm, we select candidates for our hardware acceleration. We design our hardware accelerators of the chosen candidates. The results show that our implementations achieve speedups as high as 60 folds for specific functions and 5.4 for the overall algorithm compared with the performance of the software-only implementation. Through the associated hardware cost analysis, we point to an opportunity to perform these functions in an SIMD fashion.