Side-channel Analysis And Evaluation Technology Research Base On Power Leakage

As the base of the secure information system, Cryptography is very important. However,the security of the cryptographic algorithms design is not equivalent to that of the implementation of cipher chip. There exist security risks of physical leakage during the operation of the cipher chip. Without interference the chip, the attacker can obtain the time,power consumption, electromagnetic radiation, and so on, and will be used to analyzed according to these unintentional leakage, which called 'side channel attacks' or 'bypass attack'.The side channel attack and defense strategies have become important branches of cryptography research, and are taken the attentions by physical security ?smart card chip ?measuring technique and microelectronics, etc.The characteristic of side channel attacks is using cryptographic algorithm side channel information to attack physical implementation, and find the information of security key through the non-prevented physical information leakage of black box model. In this paper, a typical side channel attack technique is studied for the physical leaks generated by the block cipher algorithm on the cryptographic chip, and the improved side channel analysis method is designed to improve the applicability of the attack and to extend the applicability of the attack. Combined with the specific security of the cryptographic algorithm under the black box model, the evaluation model of the cryptosystem is constructed. Moreover, we analyzed the relationship between security and side information, which is theoretical exploration and algorithm validation of the side-channel security problem. The main innovations as follows.1. In the template attack, using the information of power trace, choose the characteristics reasonably, enhance template attack effect by improving the matching capacity. This paper proposes a method for extracting the characteristics of power traces based on ESN, which has the advantage of simple training algorithm and comprehensive optimization capability, and then extract the characteristics of power traces.(1) This method realizes the nonlinear feature extraction of the power trace,and reduces the influence of power traces caused by non-strict alignment in the time domain because of its good fault-tolerant capability, and improve the utilization rate of side channel information.(2) The parameters selection problem of reservoir in classification method of ESN has been solved by time prediction precision sequence which is the standard. The feature extraction capability of ESN for power traces of a rough alignment were tested and evaluated by using the grid method to advance the optimization search of parametric space and taking advantage of the capability of neural network which takes the data samples as a quantitative knowledge to manage.(3) The tests have shown that this method can reduce the dependence caused by template attack on the pretreatment technology of power traces, and also improve the classification precision of the correct key.Concerning the clustering attacks,a small amount of labeled information is used to assist guide the clustering process of clustering unlabeled data, which reduces the assumption condition of the module attacks.(1) The machine learning technology is used to further conduct statistical analysis for the power data, and the supervised machine learning methods shows the better performance than the traditional template attacks.(2) With the help of experiments studies the impact of the feature selection and other factors on the key recovery, and analyze the dependence of power data. Comparing with the conventional semi-supervised template attack method, the characteristics of the extraction method reduces the abnormal data and noise interference effectively and improve the utilization ratio of priori information and key recovery success rate.3. In terms of the evaluation of the side channel, the relationship between security and side information is analyzed from both single point and network cryptography system, and presents an DPA analytical model based on hypothesis testing, which combining Welch t test with DPA(differential power analysis) method.(1) The model makes use of Welch's t-test to assess the leakage of side channel in the running process of the cryptographic algorithm, extracts side channel attribute from physical realization of cryptographic algorithm, establish quantitative relationship between the leakage of side channel and DPA success rate, and quantify the capability of cryptographic module against side channel analysis.(2) The side channel test data, which belongs to the process of various encryption algorithm,is used to verify this model,and analyze the relationship for T value and quantity of power traces and success rate. The effectiveness of the model is also proved.(3)The security performance of single node is investigated under the condition of information sharing,the analysis of preference among nodes is studied in the process of modeling, and the collaborative network security is evaluated from the whole view of network topology connection.