Research On High-level Attacks On Masking Algorithms Using Neural Networks

In current side channel attack(SCA)technology,power analysis attack is one of the best means of attack,and its development has become increasingly mature.SCA is based on the information which leakage of the energy consumption during the operation of the encryption electronic equipment,and then attacks and obtains the secret information such as the round key of the encrypted device.The effectiveness of the attack is far higher than the key analysis technology which is based on the mathematical method.Therefore,it poses a serious threat to the current encryption equipment.On the other hand,the defense measures against this attack mode have been gradually applied to the encryption process of cryptographic devices.In these defensive ways,mask strategy has obvious advantages in many aspects,such as cost,feasibility,resistance,and so on.Therefore,it has also become the focus and hot in SCA defense technology.So,the power analysis attack against mask strategy has also become a major problem in the field of SCA.The mask strategy can resist first order attack to some extent,and in many instances,it is proved that there is no first order leakage problem in the encryption process after the use of the strategy,but it may be attacked by high order power analysis to invalidate its resistance mechanism.Although the high order power analysis can attack the key information in the masking implementation to a certain extent,there are many shortcomings in the existing technologies about the high order power analysis attack such as the time complexity and the feasibility of the algorithm,both at home and abroad.So,it is still a major problem that about how to carry out the encryption process for the masking implementation.And it needs to be explored and improved.In this study,a high order template attack based on neural network algorithm is studied for the AES encryption process in cryptographic devices,aiming at the AES encryption algorithm which is masking the implementation.The goal is to propose a new idea in the high order attack,to optimize the shortcomings and defects of the existing high order attacks at home and abroad,and to achieve a higher attack algorithm to achieve the feasibility and attack success rate under the more harsh attack conditions.This will further enhance the security of the cryptographic devices,such as smart cards,in the field of hardware security.The study summarizes the current domestic and foreign development status and the existing technology in the high order attack,explores its defect and its root in theory.And it proposes the high order template attack based on neural network,which has self-adaptive high order power analysis attack.And through the theoretical basis of power leakage model,high order DPA attack and high order template attack in the power analysis attack,the corresponding mathematical model is established for the high order template attack based on neural network algorithm,and then the attack calculation method can be used in the practical application.After the implementation of the algorithm,we will use the AES encryption algorithm in the masking implementation to test the attack effect of the attack algorithm,and further analyze and study the experimental results.In this study,using the high order template attack based on forward fitting neural network,10000 energy trace is used as the experimental data for the AES encryption algorithm in the encryption device.And it is divided into the training set and the test set.The training set contains 8000 energy traces,and the test set contains 2000 energy traces.In addition,this experiment train and attack energy traces in two ways that used multiple points of interest points and interest intervals.The efficiency of the attack algorithm is quantitatively evaluated by comparing the result of training and the effect of attack.Through the experimental results,we conclusion that neural network which can fit the nonlinear relationship and have high adaptability can effectively compensate for the shortcomings and defects of the current pin for the mask strategy attack,including high order DPA attack,high order template attack and so on.The high order template attack based on forward fitting neural network in this experiment is better than the common high order DPA attack algorithm in realizing the difficulty,the attack success rate and the time complexity of the algorithm,which proves the feasibility and the research value of the attack way and the attack thought in the actual SCA domain.And the attack mode does not require the training set of known mask,which greatly reduces the requirement of the attack condition: it does not need to find the clear leakage position in the energy trace before the attack,which is the greatest advantage of the attack mode.In addition,neural network can automatically find a more suitable combination of intermediate values of power leakage models,which is another great advantage of the attack algorithm.At the same time,the algorithm is compared by using two kinds of POIs sampling method,which makes the algorithm more flexible in the use of the algorithm.