Overcoming significant noise: Correlation-template-induction attack

Due to low Signal to Noise Ratio (SNR) in general experimental environments, previous attack methods such as correlation power analysis (CPA) do not always screen out the correct key value. Sometimes the success rate of the attack is so slight that we have to find other ways to make certain of the prosperity. In this paper, rather than adopting the traditional means of singling out a single key value, we suggest a way of setting up a threshold for the attack. Accordingly, we propose a feasible method to filter the inherently enlarging candidate key space, which is called correlation-template-induction attack. The method contains three steps: First, we apply a variation of CPA and get a set of candidate key values. Then, we filter the candidate key space with template attack, which is easy to implement and requires encryptions of just a few input data to screen out the correct key. Next, to achieve optimal of our attack, we mix the concept of induction together with our attack. The experimental results given in this article on an AES smart card implementation guarantee the effectiveness of our method.