Adaptive Chosen-Plaintext Correlation Power Analysis

Yongdae K ea al. poposed biasing power traces to improve correlation in power analysis attack in 2010. However this method abandons large numbers of power traces which is unreasonable in comparison with traditional CPA. In this paper, the traces acquirement process is divided into two stages. In the first stage, some plaintexts are chosen randomly and two most probable key byte candidates are recovered. In the second stage, we adaptively choose specific plaintexts corresponding to the traces with high signal-to-noise ratio, encrypt them, and acquire the second batch of traces. So the attack can be finished with fewer traces. According to our experiments on AT89S52 software implementation of AES, getting the same success rate 0.955, our adaptive chosen-plaintext CPA only requires 78.9% traces of traditional CPA. Our proposal can be implemented by automatic software through two interactions with the AT89S52.