Research On Rogue Behavior Detection For Android Applications

Users usually obtain Android applications through application markets to meet the needs of life and entertainment.There are a large number of third-party application markets that provide application download services for Android users.In order to win users,application markets usually conduct strict scrutiny of applications entering the market,so as to allow users to obtain safe applications as much as possible.In order to improve users' experience,major application markets urgently need an automated solution to avoid some rogue behaviors that affect users' experience in applications,such as rogue advertisements that induce users to click and download applications,rogue pop-up boxes that cannot be closed normally,and rogue floating windows that affect users' experience.At present,application markets cannot effectively find rogue advertisements,pop-up boxes and floating windows that affect users' experience.To address such problems,a rogue behavior detection framework is proposed.First,the depth-first traversal strategy is used to run the application to record the information of the application interfaces and the information of the events leading to the interfaces' transition,so as to obtain the state transition diagram of the application interface.Then,the target detection method is innovatively used to identify the advertisements in the interface,and the random forest classifier based on the decision tree is used to identify the pop-up interface.Finally,combining optical character recognition technology,natural language processing technology and heuristic rules to realize the identification of rogue advertisements,based on heuristic rules to realize the identification of rogue pop-up boxes,combining image pixel analysis and heuristic rules to realize the identification of rogue floating windows.The proposed framework can also record the evidences of the rogue behaviors in applications,so that application markets can ask developers to rectify.A total of 4000 Android applications were crawled from a third-party application market named Mango Download Station.Based on heuristic rules,118 Android applications containing rogue behaviors were found by the proposed framework.The precision of rogue application detection reached 96.7% and the recall rate reached 90.1%.