Improved Differential Power Analysis Attacks on AES Hardware Implementations

This paper presents an alternative way to enhance power analysis attacks on AES hardware implementations for wireless sensor network (WSN) nodes. The proposed attack method adopts hamming differences of intermediate results as the power model and arranges plaintext inputs to maximize the differences of power traces. A simulation-based experimental environment is built, and various power attacks are conducted on our AES hardware implementation. Unlike on software implementations, conventional power attacks on hardware implementations may not succeed or require more computations. However, our proposed method improves the success rate effectively using acceptable number of power traces and fewer computations. Furthermore, experimental results also demonstrate that the linear operations of AES hardware implementations extremely leak the data-dependent power information vulnerable to power attacks.