Design And Research On Distinguisher Based On Colony Algorithm

In the area of data mining, clustering analysis is an effective way to extract information from huge data set. Clustering analysis is employed in various fields for useful information extraction for support. Ant colony algorithm simulate the process of hunting food for finding the shortest path. Molding the phenomenon, researchers obtain new algorithm with robustness for handling huge scale complex problem.In the thesis, we firstly summarize the clustering analysis. We introduce some clustering methods and similarity rules in detail. We also analyze clustering methods based on intelligent algorithms and its applications. So, we introduce two ants based on clustering models, they are ants foraging model and LF algorithm. We present the two model in detail and propose that we can use these model in power analysis area.Power analysis is important subject of information security. In this thesis we summarize the history of power analysis. We introduce the principle of power and distinguishers, such as DPA(Differential Power Anlysis),CPA(Correlation power analysis),MIA(Mutual information Anlysis),etc. For better applications, we introduce power analysis theory. Alos, we analyze interrelation between distinguishers for application for ants based clustering.We implement the AES-128 algorithm with software and obtain the power traces when the process is run. We propose the main idea of the distinguisher construction. Power leakage is dependent on the intermediate values, same value cause the same leakage. Our methods firstly cluster the power trace with some leakage. Then, we can calculate the characteristic of each class. Though our experiments, we find that even if in a bad probability of clustering successful rate we can also get the true secret key with 100% rate which illustrate that our method is resistant to noise to some extent.The method proposed in the thesis can be refined with some leakage model, such as hamming weight model. If adversary use the hamming weight when clustering, we can obtain fewer classes get better improvement of clustering. Also, our method also can combine multi leakage points compare to DPA or CPA.Finally, we summarize all our work and look into the future.