Hardware Implementation of SM4 Based on Composite Filed S-box and It's Security against Machine Learning Attack

摘要

In this paper, we implemented the SM4 block cipher algorithm based on both composite filed S-box and look up tables (LUTs) S-box. We also explored the performance of SM4 against machine learning attack, including conventional classifiers, SVM for example, and convolutional neural network (CNN). Implementation of SM4 based on composite filed S-box has a great advantage on area consumption compared with conventional implementation based on LUTs, which can be widely used in resource constrained applications. On the other hand, power attack based on CNN poses a serious threat to the security of block cipher using S-box in LUTs, while experiment show that SM4 using composite filed S-box has a better security against linear classifiers and CNN attack.