Research On Side Channel Attack Method Based On Cryptographic Chip

As the core of cryptographic algorithm,cryptographic chip is used to protect private data.It is often used in ATM,bank cards,servers,smart cards and other devices.In the research on the security of cryptographic chips,in addition to the research on the flaws and vulnerabilities of encryption algorithms,side channel attack also become the focus of the majority of scholars.This attack method can easily recover the secret key by using the leaked physical information of cryptographic chips.Its attack effect is remarkable.The research on the side channel attack of cryptographic chip is a kind of risk assessment,which is conducive to the improvement of encryption algorithm and the optimization of the internal structure and security performance of the chip.This paper constructs a side channel information acquisition system,optimizes the alignment method of side channel information,and proposes the optimal acquisition strategy.Aiming at Advanced Encryption Standard(AES)and RC4 encryption algorithm,four kinds of attack methods are optimized,which are electromagnetic analysis,power analysis,template attack and machine learning.The main work and innovation of this paper can be summarized as follows.1)Aiming at the time offset problem of captured side channel leakage signals,a method of alignment of side channel information is proposed to enhance the attack effectiveness.The correlation alignment method uses the correlation coefficient of two curve segments to realize the curve alignment.The method has the problem of local maximum correlation coefficient,which makes the curve locally aligned.Therefore,a bimodal correlation alignment method is proposed in this paper.One segment of the bimodal is used to determine the translation of data,and the other segment is used to monitor the alignment effect of the curve in the second half.The experimental results prove that the correlation alignment method has the problem of local alignment and the feasibility of the new method.From the comparison of the attack results before and after data alignment,it can be seen that the attack effect of the latter is improved in two aspects:the Spike amplitude and success rate of key guessing.The Spike amplitude of the correlation electromagnetic curve increased by 75%.Meantime,the success rate of a key of a certain byte changed from 0 to 100%.The influence of different measurement strategies on electromagnetic analysis is tested and verified theoretically.2)Taking the traditional AES encryption algorithm as the attack target,optimize differential power analysis(DPA)attack scheme based on different leakage models.In single-bit DPA attack,the influence of the information leaked by different bits on the attack is analyzed,and the best attack scheme is determined.Then a new design scheme of multi-bit model is proposed,which is called seven-bit model DPA attack method.Meantime,the feasibility of the algorithm is verified from theoretical and experimental aspects.This method effectively utilizes the leaked information of multiple bits and effectively improves the attack effect.Compared with single-bit DPA attack,the Spike amplitude of the new method is increased by 101%.3)AES with random progress interrupts(AES-RPIs)encryption algorithm increases the noise of side channel signal,which weakens the effect of side channel attack.To solve the problem that the frequency of noise is uncertain,this paper presents a correlation power analysis(CPA)attack method based on empirical mode decomposition(EMD).In the new method,the complex signal is decomposed into several local characteristic signals of different time scales and used as the actual leakage model to implement CPA attack to enhance the attack effect.Meantime,an attack effect evaluation scheme is proposed,which uses Spike visibility to amplify the Spike effectively.The experiment proves that the attack effect of the new method is significant.The Spike contrast of the new method is 64%higher than that of traditional EMD and 3.46 times higher than that of traditional CPA attacks.4)Taking AES-RPIs encryption algorithm as a target,a blind Sliding window CPA(SW-CPA)attack method is proposed for the extreme cases where the insertion position of random interference is unknown and the Spike of difference curve or correlation curve is unknown.Without reference to the Spike position of the curve,the method directly aggregates the information of the scattered original power curve.At the same time,the form of traversal sliding window is adopted to ensure the best superposition effect,which is more in line with the actual situation.Two sets of data are used to analyze the attack effect of the new algorithm.In the attack results of the first set of data,compared with CPA,the Spike value increase range of the new method is 2.98-7.32 times,compared with SW-CPA,the increase range is 1.96-5.25 times.In the attack results of the second set of data,compared with CPA,the Spike value of the new method increased by 79.7%;compared with SW-CPA,the new method increased by 29.8%.And the number of curves consumed by the new method is reduced by 15%when the attack success rate is 100%.5)In the attack against AES-RPIs encryption algorithm,this paper proposes SW-CPA method based on compressed energy entropy.In this method,the scattered information is superimposed by sliding window.And the curve is divided into several segments to obtain the energy entropy after compression,which makes the effective information focus and weakens the influence of random progress interrupts(RPIs).The new method can effectively reduce the influence of random interference on the leakage rule of instruction operation in the encryption algorithm.Firstly,the feasibility of the algorithm is verified by theoretical derivation.Then,the attack effect is verified by experiments from data complexity and Spike amplitude.In the analysis of the first set of data,compared with the CPA attack,its Spike amplitude increases by 3.89-9.51 times,and by 2.64-5.89 times compared with SW-CPA.When the attack success rate was 100%,the number of curves consumed by the new method was 35%lower than the other methods.6)In the research of template attack based on Bayesian theory,two screening methods were proposed to further screen the interest points to build the template.This paper studies the impact of the data amount of template construction,the data amount of template matching and the selection of interest points on the attack effect,so as to determine the best template construction and matching scheme.In terms of machine learning attack,a hybrid method of Independent component analysis(ICA)and principal component analysis(PCA)is proposed,which makes full use of the characteristics of side channel information and weakens the high order correlation between data,and improves the prediction success rate.The experimental results show that in different dimensions,compared with PCA method,the increase range of classification success rate of ICA method is 11.1%-230%,and the increase range of fusion algorithm is 33.3%-230%.In the same case of data consumption,the prediction success rate of fusion algorithm increased by 6.69 times compared with PCA,and that of ICA method increased by 5.15 times compared with PCA.