Research On Android App Third-Party Library Detection And Privacy Leakage

With the development of the Android system,the increasingly complex functional requirements increase the workload of application developers.To speed up the application development process and obtain advertising revenue,developers often introduce third-party libraries into applications,but third-party libraries may bring privacy leakage problem.Therefore,it is of great significance to detect and analyze the privacy leakage of thirdparty libraries.There are two types of tools for third-party library detection:clustering-based tools and similarity comparison tools.However,the existing clustering-based tools still use accurate feature matching,which is difficult to deal with code obfuscation such as code removal and control flow randomization.Besides,the tools based on similarity comparison cannot detect closed-source third-party libraries.The dynamic analysis methods for privacy leakage include network traffic analysis and dynamic taint analysis.The former ignores various interfaces such as logcat,while the latter requires modification of system code,which is inconvenient for migration and adaptation.To solve these problems,we propose and implement a novel tool named libmonitor.First we classify Android APIs,and use the call frequency of APIs in each category as the feature.We perform a cluster algorithm to detect third-party libraries through fuzzy matching.Next we use dynamic instrumentation technique to monitor privacy sensitive Android APIs.After matching the encoded sensitive information and analyzing the call stack,a privacy leakage analysis report of the third-party library is generated.In this thesis we evaluate third-party library detection on 162 groups of obfuscated applications and 217 real-world applications,and compare with two existing third-party library detection tools based on clustering.The experimental results show that the F1 values of libmonitor in two sets of experiments are better than the other two tools,and libmonitor found 5809 pieces of privacy leakage risks involving 152 third party libraries through dynamic analysis of 64 applications.