Analysis Of Usage Patterns And Privacy From Massive Android Usage Data

The Android Market,with its good user experience,the sheer number of apps and the simplicity of the installation process,makes it a popular way for users to get mobile apps.Meanwhile,it has also becomes the primary target of malicious attackers and the main spread platform for malware.Current categorization in the Android Market studied do not exhibit a good classification quality in terms of the claimed feature space.On one hand,they are inherently unresponsive and un-adaptive,and therefore cannot speedily adapt to shifting developer behaviors and market preferences,nor can it help to identify emerging technical trends.On the other hand,these categorization approaches are theme-based which may fail to explain an app's functionality,and do not cluster apps according to the features they exhibit.An effective categorization of software according to its functionalities offers advantages.On one hand,an approach that can automatically characterize the different types of applications can be helpful for both detection fraudulent and organizing the Android Market.On the other hand,such categorization can also help app developers by facilitating code-reuse,locating desirable features and technical trends within domains of interest.The existing categorization methods are based on static features of app software,such as language features and structural features.In order to improve the categorization accuracy,we propose an app categorization method based on apps' usage patterns.In the age of big data,people are increasingly concerned about personal privacy leaks.However,the growing popularity of smartphones and offer ubiquitous mobile connectivity have also increased concerns about the privacy of users who operate these devices.These concerns have been exacerbated by the fact that it has become increasingly easy for users to install and execute third-party applications.In this paper,we study the privacy threats that applications,written for Android,pose to users.Our work is summarized as follows:(1)We analysis the current static method of classification feature extraction for mobile applications.They use static features to develop feature models,such as,extracting software features from an application's textual artefacts,API calls,system permissions from the application's XML files and user's behavior features.In this paper,the shortcomings of static characteristics are analyzed in detail.(2)We propose an app categorization method based on apps' usage patterns.Through our experiments,it is observed that the main function of apps and their usage patterns are correlated,based on that we could categorize apps more accurately.The results showed that the proposed method is superior to the methods based on the static features,which effectively improves the quality of categorization.(3)In this paper,we analysis personally identifiable information leaks in mobile usage data.We found that there are more than 14 types of privacy information that leaks in user's usage data,meanwhile,user's habits and preferences in usage data.