摘要
近年来,深度学习技术广泛应用于侧信道攻击领域.本文提出了一种使用改进残差网络和数据增强技术,直接恢复密钥字节的能量分析攻击方法.首先将残差网络模型中的二维卷积核改进为适用于处理能量迹的一维卷积核,然后使用数据增强技术对原始能量迹增加高斯噪声和随机相位抖动,最后使用训练好的模型实现密钥恢复.通过实验对现场可编程逻辑门阵列(FPGA)实现的AES分组密码算法进行了攻击,使用“分而治之”的策略,对128比特密钥的最后8比特进行了恢复,该模型平均测试精度达到65.48%,与卷积神经网络(CNN)和多层感知器(MLP)神经网络相比,同等条件下测试精度分别提高了16.63%和54.27%,并在ASCAD公开数据库上对模型的性能进行评估.该模型使用数据增强技术解决了小样本训练问题,减少了训练过程中过拟合现象的发生,模型对噪声和相位抖动具有很强的鲁棒性,通过改进结构有效减少了模型参数和节省了计算资源,为密码芯片能量分析攻击提供了一种新的思路.
In recent years,the techniques of deep learning have been widely used in the field of side-channel attacks.This paper proposes a power analysis attack method using an improved residual network and data augmentation technique to directly recover some key bytes.Firstly,in order to fit the processing power traces,the two-dimensional convolution kernel in the residual network model is improved to a one-dimensional convolution kernel,and then data augmentation technology is used to add Gaussian noise and random phase offset to the original power traces.Finally,the trained improved residual network model is used to recover the key.As a practical application of the proposed method,the proposed attack is applied to the AES algorithm implemented on a field programmable logic gate array(FPGA).The last 8 bits of the 128-bit key were recovered using the“divide and conquer”strategy.The average test accuracy of the model is 65.48%,which has been improved by 16.63%and 54.27%respectively compared with the convolutional neural network(CNN)and multilayer perceptron(MLP)neural network in the same conditions.The performance of the model is evaluated on the ASCAD public database.The model uses data augmentation techniques to solve the problem of small sample training,reducing the occurrence of overfitting during training.The model is robust to noise and phase offsets.The model parameters are effectively reduced and the computing resources are saved by improving the structure,which provides a new idea for power analysis attack.
作者
王恺
严迎建
郭朋飞
朱春生
蔡爵嵩
WANG Kai;YAN Ying-Jian;GUO Peng-Fei;ZHU Chun-Sheng;CAI Jue-Song(Strategic Support Force Information Engineering University,Zhengzhou 450001,China)
出处
《密码学报》
CSCD
2020年第4期551-564,共14页
Journal of Cryptologic Research
基金
河南省网络密码技术重点实验室开放基金(LNCT2019-S02)。
