神经网络结构在功耗分析中的性能对比

Comparative study on neural network structures in power analysis

为了探究不同的深度神经网络运用在功耗分析攻击中的性能差异,在DPA_Contest_V4数据集的基础上进行实验。破解循环掩码后,首先将深度神经网络与传统的SVM等机器学习算法模型进行对比,然后分析神经网络模型结构的变化对实验结果的影响,最后结合循环神经网络,将不同的网络模型进行综合比较。实验结果表明,在相同实验条件下,神经网络模型要优于传统的机器学习模型,循环神经网络模型要优于深度神经网络模型,其中,不同层数的神经网络采取的激活函数不同,会导致实验结果发生较大变化。

In order to explore the difference of different neural networks used in power analysis,we use DPA_Contest_V4 dataset to complete our experiment.After the mask is cracked,the deep neural network and the traditional machine learning like SVM are firstly used.Then,the impact of the changes in the structure of the neural network model on the experimental results is analyzed.Finally,a comprehensive comparison of different network models is made by combining the cyclic neural network.The experimental results show that the neural network model is superior to the traditional machine learning model and the recurrent neural network model is superior to the deep neural network model when experimental conditions are the same.Among them,the activation functions of neural networks with different layers are different,which will lead to great changes in the experimental results.