Correlation Power Analysis For AES Based-on Principal Component Analysis

IOT and Internet+ rapid development makes more and more electronic devices such as smart appliances interconnection, wireless sensors, engineering machinery. The side channel attack by analyzing encryption algorithm running leak, such as time, power consumption to decipher their keys and get permission. IOT lightweight device because of its simple structure, single function more vulnerable to side channel attacks, depth study of side channel attacks is necessary. In the side channel attacks having simple to implement universal strong advantages related to correlation analysis attacks are widely used.However, by analyzing the correlation coefficient and the power point position of the correlation analysis of the AES algorithm, found that:In the power curve, there are some key points which are high correlation between the redundant point and the key low correlation zone. These points are useless to the analysis of the key. That correlation power analysis of a lot of useless points. The high redundancy point of the key correlation is that there are a number of key high related information in the high correlation area. In these points, one can analyze the key, that is, the other points are redundant. The point in the low correlation zone is that the correlation between the point in the region and all the key is relatively low.In view of these two questions, use the principal component analysis to carry on the pretreatment to the power consumption curve. Extract the highly correlated features and separate the low correlation zone. Using the maximum extraction method in curve compression, the two period is used to extract a maximum value to replace all points in the two period to reduce the redundancy. These two methods achieve the goal of reducing the amount of data analysis by reducing the power consumption curve.Experimental results show that: The principal component analysis of the power consumption data pretreatment can be all the correlation strong main component extracted before 111 principal components. Use only high correlation principal component can obtain all the key value, and analysis of the amount reduced by about 3/4.Compared with the maximum extraction method using different thresholds, the results showed that the threshold value of the two period was better, and more in line with the characteristics of MOVX. The first maximum extraction, then the principal component analysis method to better curve pretreatment, to analyze the actual amount of reduced to at least 5.5% of the original method.Through the analysis of algorithm realization, microprocessor architecture and collection environment, the reasons of these two problems are found out: For more accurate acquisition of power consumption information, flat sampling rate several times to the clock frequency, namely a clock we collected a number of power points this leads to redundant power point; The microprocessor executes non MOVX operating power consumption little will also be collected, the correlation is very low, resulting in low correlation region.