Cluster Analysis Of Malware Based On Ant Colony Optimization

Smartphones have become ubiquitous around the world,with Android phones accounting for three-quarters of all smartphones.Over time,android malware has proliferated.Most of the existing malware identification schemes are classified by supervised learning method,which has high accuracy and high speed,but poor classification effect in the face of new samples.Unsupervised learning clustering algorithm is better in dealing with unknown samples.In this thesis,aiming at the problem of poor performance of classification algorithm in processing unknown samples,a hierarchical clustering algorithm based on improved ant colony algorithm is proposed by referring to maximum and minimum ant colony algorithm,ant sorting algorithm and hierarchical clustering algorithm.Experimental results show that when dealing with malicious software data set,the clustering effect is higher than that of the hierarchical clustering algorithm optimized by the original ant colony algorithm,and the convergence speed of the algorithm is faster.The tree graph of the clustering process can be generated while the high-quality clusters are generated.The main content of the thesis includes:(1)Android malware features are extracted from both static and dynamic directions at the same time.In view of the problem that the extracted data feature dimension is too high to be used for clustering,fuzzy hash algorithm is used to process the data before clustering.(2)Based on the hierarchical clustering algorithm,the malware data clustering class was analyzed.Aiming at the problem that the hierarchical clustering algorithm could not sublimate the local optimal solution to the global optimal solution and thus fell into the local optimal solution,an improved algorithm was designed by referring to the ant colony algorithm.Experiments on a variety of data sets show that the clustering accuracy of the improved algorithm is 18.2% higher than that of the original hierarchical clustering.(3)Aiming at the problem that the improved clustering algorithm has slow convergence speed and still falls into local optimal solution in the face of complex data sets,a reimproved algorithm is proposed by using Max/min ant colony algorithm and ant sorting algorithm.the experiment shows that compared with the improved algorithm,and improved algorithm improves the convergence speed of 19.4%,in the face of complex data set clustering precision is increased by 3%.