Abstract: For the electromagnetic radiation leakage during the working of the cryptographic chip, we propose a feature selection method based on the Hamming weight model, which uses the fact that the electromagnetic radiation energy depends on the Hamming weight of the processed data. First, we construct different classifications for the different Hamming weights of the intermediate values of the encryption algorithm. Each classification uses its own selected plaintext to perform repeated encryption operations, and collects the electromagnetic signals radiated by the chip. Then, we calculate the mean value of the electromagnetic signal in each classification and the difference of the mean value of the electromagnetic signal between any two classifications to obtain a difference matrix. Finally, we define a set of variables to constrain the difference matrix, and the points that meet the constraints will be used as feature points. The experiment results show that the Hamming weight model not only can correctly select feature points, but also can significantly reduce the number of electromagnetic leakage traces required for template construction compared with the correlation coefficient feature selection method.

Key words: template attack, side channel attack, feature selection, Hamming weight model, correlation coefficient, euclidean distance, SM4 block cipher

摘要： 针对密码芯片运行过程中的电磁辐射泄漏，提出基于汉明重量模型的特征选择方法，该方法利用电磁辐射能量依赖于被处理数据汉明重量这一事实.首先，对密码算法中间值的不同汉明重量构建不同的类别，每个类别使用各自的选择明文执行重复加密运算，同时采集芯片辐射的电磁信号；然后，计算每个类别内电磁信号的均值以及任意2个类别间电磁信号的均值之差，得到一个差值矩阵；最后，定义1组变量对差值矩阵进行约束，将满足约束条件的点选为特征点.实验结果表明：基于汉明重量模型不仅能够实现特征点的有效选择，与相关系数特征选择方法相比，该方法还能显著减少模板构建所需的电磁泄漏迹数量.

关键词: 模板攻击, 侧信道攻击, 特征选择, 汉明重量模型, 相关系数, 欧氏距离, SM4分组密码