Research And Implementation Of Side Channel Attack Based On Symmetric Encryption Algorithm

The security of cryptographic algorithms is the most concerned thing.It mainly includs the logical integrity at the level of mathematical principles,and the speed of encryption and decryption.The purpose is to ensure that cryptographic algorithms are fast encrypted and difficult to be attacked.As encryption algorithms are applied to a large number of encryption devices,their security will be threatened by Side-Channel Analysis(SCA).Side-channel attacks use the bypass information leaked during the operation of encryption devices.The information contains the key information of the device,and this type of attack is hard to guard against.This thesis is mainly aimed at the research and implementation of the side channel attack method of the symmetric encryption algorithm(Advance Encryption Standrad,AES).The main work content and results are as follows:This thesis first introduces the principle of AES encryption,then introduces the principle of side-channel attacks and commonly used energy leakage models,and introduces traditional side-channel attack methods.Through the analysis of traditional methods,it is found that there are some limitations,but the current powerful machine learning methods overlap with side-channel attacks,so this type of method is applied to side-channel attacks.In this thesis,the Chipwhisperer platform is used as the experimental data acquisition platform to preprocess the collected energy consumption signals.Through the analysis of the components of the energy curve,the construction of the differential energy trace is proposed,which can remove the redundant components in the energy trace.Secondly,for the various noises in the energy trace,a method of constructing a denoising autoencoder is proposed.The energy consumption signal is filtered and the template attack experiment is carried out.Compared with the traditional wavelet denoising method.The experimental results show that the energy trace preprocessing method proposed in this thesis has a better noise reduction effect,and the signal retains more useful information.This thesis analyzes and studies the traditional template attack,and compares it with the support vector machine(Support Vector Machines,SVM)classification method.The experiment found that although the SVM is more efficient than the traditional template attack,there are still some shortcomings.Through detailed analysis,it is found that the energy consumption matrix has ill-conditioned problems.So a two-point joint coefficient feature extraction method is proposed and compared with the Sum of Difference(SOD)and Principal Component Analysis(PCA)methods.At the same time,the Multilayer Perception Network(Multilayer Perceptron(MLP)is applied to side-channel attacks,and a two-point joint feature extraction MLP network attack is proposed.Compared with traditional methods,it can achieve higher attack efficiency.This thesis also considerd the AES mask scheme,and applied the MLP network model to the side channel attack of AES mask defense.Aiming at the data set ASCAD of the AES-128 algorithm implemented by masking,the median value of the attack point was analyzed.By establishing the corresponding Hamming weight model,the MLP network is used to model and classify the data set.The results show that the attack method can successfully attack the data with mask protection.