Research on Feature Selection Algorithm of Energy Curve

Energy analysis attack is a side channel attack, which collects and analyzes the power leakage information in the operation process of cryptographic chip, and then recovers the correct key. In the process of energy analysis attack, the collected power leakage information has many feature dimensions and a large amount of data. Putting all the features into the algorithm will bring dimension disaster. Therefore, choosing the characteristic points of the energy curve is of great significance for the success of the attack. Firstly, three kinds of feature selection methods are studied in this paper. Secondly, three energy curve feature selection algorithms are implemented: dynamic feature selection algorithm based on mutual information, feature selection algorithm based on decision tree and feature selection algorithm based on recursive feature elimination. Finally, the three feature selection results are tested and evaluated by machine learning, which shows that the subsets generated by the three algorithms have good performance and can be used for energy analysis attacks. Among the three methods, the feature selection algorithm based on decision tree has a short-time and the selected feature subset is the best.