针对密码芯片数据搬移能量曲线的机器学习攻击

Machine Learning Attack to Power Traces of Data Movement in Cryptographic Chip

机器学习和传统侧信道攻击技术中的模板攻击有类似的处理过程,它们都由学习和测试两个阶段组成。模板攻击可以看作是有监督学习的分类技术,而机器学习领域也有很多这样的分类算法。为了探索机器学习算法在侧信道攻击中的应用,以实际密码芯片中的数据搬移操作为攻击对象,研究了一些机器学习算法利用已知搬移数据值的能量曲线,对新的能量曲线的搬移数据值的预测效果。结果表明,在采用一条能量曲线进行测试时,一些机器学习算法比模板攻击预测的正确率更高。

Machine learning and template attack in the traditional side channel attack techniques have similar procedures, they all consist of two phases: learning and testing. Template attack can be considered as a classification technique for supervised learning, and there are many such classification algorithms in the machine learning field. In order to explore the application of machine learning algorithms in side channel attack, using the data movement operation in actual cryptographic chip as the attack target, the forecasting effect of some machine learning algorithms is investigated. These algorithms make use of power traces with known value of the moved data, then predict the value of the moved data for some new power traces. The results show that, when employing only one power trace in the testing stage, some machine learning algorithms have higher correctness rate than template attack.