Analysis Of AES Electromagnetic Leakage Signal Based On FPGA

With the rapid development of the information industry,cryptographic equipment has penetrated into all walks of life,not only in daily life,but also in the fields of national defense,politics,finance and other fields.At the same time,the information security of cryptographic devices has received attention from various sources.During the operation of the cryptographic chip,a change in the state of the circuit produces a current,and a varying current produces a varying magnetic field.By analyzing the magnetic field information that is inadvertently leaked,the core data information of the cryptographic chip-the key can be obtained.In this paper,the classification of cryptographic chip electromagnetic leakage signals to identify different keys is of great significance for evaluating the information security of cryptographic chips.There are three main innovations:Firstly,this paper builds a fully automatic electromagnetic leak signal acquisition platform consisting of Sakura-G development board,electromagnetic induction probe,amplifier,oscilloscope and computer equipped with Checker and Transporter program.Through the platform,the electromagnetic leakage signal of the packet encryption algorithm AES during operation is collected,which saves the labor cost,and also makes the experimental environment for collecting signals to be the same.Secondly,this paper designs an electromagnetic leakage signal classification algorithm combining empirical mode decomposition algorithm(EMD),signal energy feature and gradient lifting tree(GBDT)method.Firstly,the original signal is segmented by time window to obtain the sub-signals corresponding to each round of AES encryption algorithm.Then,the sub-signals are decomposed by EMD method,and the energy features of the decomposed feature mode function(IMF)are composed into the feature matrix,and then based on the feature matrix,Gradient Lifting Tree(GBDT)method is used to realize the classification and identification of different key electromagnetic leakage signals,and the change of recognition accuracy is compared when the number of decision trees and parameters are different.By comparing with the method of principal component analysis(PCA)to extract signal features,experiments show that the designed algorithm improves the classification accuracy.Thirdly,this paper proposes to apply the deep learning network Inception ResNet V2 to electromagnetic leakage analysis,and use the Inception ResNet V2 network with different network layers and different convolution kernel types to classify the electromagnetic leakage signals,the change of recognition accuracy of network parameters is studied,and the optimal network parameters are selected accordingly.The experimental results show that the recognition accuracy of Inception ResNet V2 network is better than that of gradient lifting tree.