Side Channel Attack on SM4 Algorithm with Deep Learning-Based Analysis

SM4 is a Chinese symmetric algorithm, which is a little bit similar to AES. If the key of the algorithm is recovered, adversaries can easily make a fake message by replacing part of the ciphertext. In addition, adversaries can learn the secret message just by decrypt the ciphertext he got. Thus, it is critical to recover the secrete key. Gennerally, there are several ways for choice, like the traditional side channel analysis (SCA) and the template attack-based way to achieve recovering the target key. We introduce a new way to recover the secrete key which innovatiely use the technique of deep learning. We use the training set to train the network for specific epoches firstly, and use the well trained model to attack the target dataset and the result shows that this method can work successfully.