Study On Android Malware Detection Method Based On Machine Learning

With the rapid development of technology and the continuous updating of Internet,the use of smart phones has become more widespread.Smartphones that use Android as the operating system already account for a large share of the smartphones market,while the openness of the Android platform also attracts many malware developers.As long as users download software that contains malicious code,it is possible to produce privacy theft,malicious deductions and other hazards.Therefore,how to detect Android malicious software efficiently and accurately is the hot spot of current research.This article has thoroughly studied the Android system architecture and security mechanism.Based on the analysis of the research status at home and abroad,the following work has mainly been done:(1)Aiming at the defect that the weight value of each attribute in Na?ve Bayesian classification algorithm is the same,Aiming at the defect that the weight value of each attribute in Na?ve Bayesian classification algorithm is the same,an improved scheme of adding weight coefficients to the feature attributes is proposed.Based on the improved algorithm,a scheme to detect Android malware is proposed.The scheme uses the permissions of the application as a feature set and optimize them,the improved Naive Bayes algorithm is used for classification testing.Experiments show that this detection scheme has significantly improved the detection accuracy.(2)Different classification algorithms have different effects on the classification of different feature sets.In this paper,a model of Android malware detection based on multi-algorithm fusion decision is proposed.In this scheme,the classification results of different types of feature sets are obtained by corresponding main classification algorithm and tuning algorithm,and the results of different types of feature sets are fused to get the final classification result.Experiments show that the detection schemes have higher efficiency and better accuracy for malware detection,the accuracy is 95%.