| The side-channel attack is a powerful cryptographic security analysis method,which mainly uses the physical information leaked by the encryption device when the encryption algorithm is running to obtain the key.Among them,the power analysis attack is widely favored by researchers due to its simple collection and the diversity of processing methods.In this research,power attack is performed on the elliptic curve digital signature algorithm,and the main works are as follows.This thesis introduces the common implementation methods of scalar multiplication in elliptic curve algorithms,such as Double-and-Add scalar multiplication,classic Montgomery Ladder scalar multiplication,etc..At the same time,it elaborates on the side-channel attack analysis methods that researchers have carried out for different implementation algorithms.Based on this,Itoh et al.proposed a Montgomery Ladder scalar multiplication aginst the address attack(Address-DPA).This proposed algorithm is the core algorithm of this thesis.In view of the shortcoming of the traditional template attacks,the research uses feature selection to optimize it.However,when using traditional algorithm such as SOD and SOST in research,it found that there are more redundant information in the selected feature subset.Therefore,this thesis proposes an improved unsupervised feature selection algorithm and uses template attacks to verify the results.The experiment result shows that the improved feature selection algorithm performs better.This research conducts an in-depth study on the traditional side-channel attack,and a recursive address collision attack method is proposed.Aiming at this method,this article analyzes the feasibility based on the experimental platform built.And the results show that the attack proposed in the research is feasible.This thesis carries out key recovery from two aspects of the horizontal side channel attack and the vertical side channel attack separately.In the horizontal side channel attack,the thesis uses K-Means clustering and Gaussian mixture clustering algorithm to conduct experimental comparisons.The results show that the Gaussian mixture clustering algorithm is more effective;in the vertical side channel attack,this research uses SVM and random forest algorithm to build classifier to recover all of the key information.The random forest algorithm composed by multiple weak classifiers,which has strong generalization and anti-interference ability.Moreover,its classification accuracy is high and is not affected by irrelevant information.Finally,using the classification results for key recovery,there is a problem that the classification accuracy is not proportional to the key recovery success rate.Based on this,this thesis proposes a new attack method which joints with multi-point,combined with the soft classification method.This method can obtain 99% key recovery success rate. |
