论文部分内容阅读
In recent years, statistical methods have been widely used in the field of SCA (Side Channel Attack), leading to a variety of SCA approaches.In this paper,we propose a novel SCA method called PRA (Pattern Recognition Analysis) by introducing the pattern recognition technique into DPA (Differential Power Analysis).The main idea of our new approach is as follows: First of all, we establish two kinds of machine learning processes, i.e., supervised learning and unsupervised learning,which can be used to extract some effective features from power data.Then based on these features we design a pattern distinguisher which can be applied to discriminate correct key-dependent classification from the wrong ones.At last, we can predict the correct key by utilizing the designed distinguisher to deal with power traces.In order to illustrate the effectiveness of PRA, we implement PRA attack experiments on unprotected circuit and circuit with countermeasure of masking.The results show that PRA, especially second-order PRA has obvious advantage over the previously known DPA approaches.