计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (6): 110-117.DOI: 10.3778/j.issn.1002-8331.2009-0397

• 网络、通信与安全 • 上一篇    下一篇

基于条件生成对抗网络的侧信道攻击技术研究

汪晶,王恺,严迎建   

  1. 战略支援部队信息工程大学,郑州 450001
  • 出版日期:2022-03-15 发布日期:2022-03-15

Research on Side Channel Attack Technology Based on Conditional Generation Against Network

WANG Jing, WANG Kai, YAN Yingjian   

  1. PLA Strategic Support Force Information Engineering University, Zhengzhou 450001, China
  • Online:2022-03-15 Published:2022-03-15

摘要: 近年来,深度学习技术广泛应用于侧信道攻击(side channel attack,SCA)领域。针对在基于深度学习的侧信道攻击中训练集数量不足的问题,提出了一种用于侧信道攻击的功耗轨迹扩充技术,使用条件生成对抗网络(conditional generate against network,CGAN)实现对原始功耗轨迹的扩充,并使用深度神经网络进行侧信道攻击。通过选择密码运算中间值的汉明重量(hamming weight,HW)作为CGAN的约束条件,将CGAN生成模拟功耗轨迹作为多层感知器(multi-layer perceptron,MLP)神经网络的训练数据,构建模型实现密钥恢复。通过实验对不同类型训练集的攻击效果进行比较,结果表明,使用CGAN生成的功耗轨迹和原始功耗轨迹具有相同的特征,使用扩充后的功耗轨迹对MLP神经网络进行训练和测试,训练精度和测试精度分别提高15.3%和14.4%。

关键词: 侧信道攻击, 深度学习, 条件生成对抗网络, 多层感知器

Abstract: Deep learning technology has been widely used in the field of side channel attack(SCA) in recent years. To solve the problem of insufficient training set in deep learning side channel attacks, an energy trace expansion technology used in SCA is proposed, conditional generate against network(CGAN) is applied to augment the original energy traces as well as training deep neural networks for side-channel attacks. The hamming weight(HW) of the intermediate value of the cryptographic operation is chosen as the label of CGAN. The simulated energy traces generated by CGAN are used as the training data of the multi-layer perceptron(MLP) neural network, and the model is trained to achieve key recovery. The attack effects of different types of training sets are compared through experiments, it shows that the energy traces generated by CGAN have the same characteristics as the original energy traces. The training accuracy and test accuracy are increased by 15.3% and 14.4% respectively with the usage of augmented traces in the training and test of MLP neural network.

Key words: side channel attack(SCA), deep learning, conditional generative adversarial network(CGAN), multilayer perceptron(MLP)