The Anaysis And Detection Of Fraud Android APPs Based On Big Data

With the rapid development of science and technology,the mobile phone has become one of the indispensable necessities in people's life.However,due to the lack of effective detection of Android mobile applications,the attendant security issues are endless.For example: many applications in mobile phones can steal the important information of Android mobile phone users in an illegal way,causing great losses to the vast number of users.The main research work of this paper are as follows:This paper presents a fraud android APP analysis and detection system for the increasingly rampant fraud android apps.Firstly,extracting the static characteristics and dynamic characteristics of android applications,and the binary feature vectors of the application are generated by combining static features and dynamic features.Then,the collected data sets are trained with the machine learning algorithm and the deep belief network in deep learning algorithm to generate the corresponding classification model.Finally,using the generated classification model to analyze the android applications.System takes advantage of the combination of static characteristics and dynamic characteristics of android applications,make up for a lack of test code coverage of static characteristics and high rate of false positives of dynamic characteristics.In this paper,15000 Android APPs were collected and tested.The experimental results show that the deep learning algorithm compared with other traditional machine learning algorithm has a higher accuracy.The correct rate is up to 97.08%.In the case of using the same data sets,the accuracy of deep belief network is 4.59% higher than random forestalgorithm,16.45% higher than support vector machine algorithm,15.33%higher than the K neighboring algorithm.