| Side channel attack(SCA)is a powerful attack method,which mainly targets cryptographic chips and cryptographic devices.It uses the relevant information leaked during data encryption to crack the encryption key used in the encryption device.SCA can obtain key information without destroying cryptographic equipment and encrypted data in the process of data encryption,which poses a serious threat to the security of cryptographic equipment.This paper mainly studies the side channel attack method and implementation scheme of Advanced Encryption Standard(AES)algorithm.The work content and research results are as follows:Firstly,it introduces the principle,classification and background of AES attacks,and then introduces the research background of AES attacks in detail.Finally,the application of machine learning and neural network in side channel attack is introduced.Through the analysis of the collected power consumption curve,it is found that the components of the power consumption curve contain noise interference,which will have a great impact on the results of side channel attack.In this paper,a hybrid denoising method is used to optimize the power consumption curve.Then,the flow of CPA is analyzed in detail.It is found that the method used to calculate the correlation coefficient between guess key and actual power consumption using Pearson correlation coefficient is calculated item by item.Aiming at this problem,a method similar to online calculation of correlation coefficient is proposed,and then the two calculation methods of correlation coefficient are compared,The method proposed in this paper takes less time to crack the key than the Pearson correlation coefficient calculation method.By analyzing the principles and steps of classical template attack,according to the problems of high overhead and low success rate in template attack,an attack method based on rough set theory and support vector machine(SVM)algorithm is proposed.The experimental results of this method are analyzed.It is found that rough set theory only reduces redundant data and does not divide data with different characteristics.According to the problems of rough set theory,a method is proposed to reduce the dimension of data set by linear discriminant analysis(LDA)and train the reduced dimension data set by SVM algorithm.By using ascad data set as the attack target,a side channel attack method of principal component analysis(PCA)applied to multilayer perceptron(MLP)is proposed.It is found that neural network has a strong attack effect on masked AES algorithm.Finally,the results of classical template attack are compared with those of the attack methods proposed in this paper,and the performance of each attack method is analyzed. |
