A Power Attack Method Based on Clustering

Abstract

Clustering uses the similarity between samples to automatically classify, and the power information generated in the encryption process has a certain similarity. Therefore, a method that use clustering for power attack is present in this paper. Similarity between the power data is considered in the method. We divided the data into many categories by clustering, and classified the data according to the guess key. We compared the similarity between the two categories, and got the correct key in the case of greatest similarity. At the same time, the s-box simulation data of AES encryption algorithm was selected as the verification example. We selected the key-related part of the power data as feature points to attack based on K-means clustering, and used Euclidean distance to calculate the similarity between clustering and classification results. The experiment result shows that the method can help us get the correct key effectively.