Research On Similarity Detection Of Android Application

The Android platform has a big quantity of applications which includes different categories.Among the whole Android market,which also exists many pirated applications.They just slightly modify the genuine applications' code and resource files.These pirated applications have brought a lot of challenges to the Android application market.Facing these pirated applications,most of detection methods have these problems:difficult to describe the characteristics,complicated extraction algorithm and poor processing efficiency.As a consequence,how to quickly detect pirated applications among the big data,and how to detect newly submitted applications,already becomes to big issues that has to be setteled urgently.In order to solve these problems,we designed one kind of detection system to screen pirated applications from the perspective of applications'similarity.this detection system based on class-directory and screenshot-interface.The main contents are as follows:(1)The analysis program based on the class-directory structure:Due to the APK file of the application,we must get all the code structure information firstly,and then change the class-directory of code into a tree structure,meanwhile specifically contrast several types of label representation,select the hash of the class name as a tree node.After that,The non-linear tree nodes through the Depth-First Travelsal algorithm,the tree node connected to a string as the signature of the application.Finally the similarity degree is compared by normalizing editing distance.(2)The analysis based on screenshot-interface,it is an image analysis of the application interface screenshots.First,we compare several common image charateristics extraction algorithms and use the SURF algorithm to collect the feature points of the image,then analyze the Android interface.Compared with the general scenery image,the background is clear and single.Therefore,the wavelet response parameters of the feature points can be set to reduce the number of feature points.And the feature points are corrected by the interval position of most feature points.Lastly,the feature point array is signed as the application.(3)Finally,for social applications on the Android platform piracy detection,we can get the correlation diagram of the similarity applications.For newly submitted application,we can accurately determine the association with the existing application.