Power Analysis on SM4 with Boosting Methods
Mengmeng Xu, Liji Wu, Xiangmin Zhang · 2018
Crypto devices may undergo attacks that beyond classical cryptanalysis. Side channel attack (SCA) gathers leaked information, such as electromagnetic radiation and power consumption to extract the secret key. Power analysis attack is one of the most efficient SCA methods. Machine Learning encompassed tools that perform smart analysis of data. This work comprehensively investigates the application of boosting method in power analysis attack. The considered methods include: Adaboost, Gradient Boosting Decision Tree(GBDT), eXtreme Gradient Boosting(XGBOOST) and Lightgbm. The boosting method is applied to exploit the power consumption of crypto device. Our promising result confirms the significance of employing machine learning method in power analysis attack. We achieved 10% on accuracy with boosting method compared with template attack.