基于数据流分析的Android应用隐私泄露检测研究

发布时间：2018-04-03 18:31

本文选题：Android　切入点：隐私泄露　出处：《南京大学》2017年硕士论文

【摘要】：随着技术的进步,智能手机给人们的生活带来了极大的便利。据统计,搭载Android智能系统的手机现如今已经占据了 81%的手机市场份额,这也正让安卓系统成为了攻击者窃取用户隐私数据的一个重要目标。因此,由智能手机所导致的用户隐私泄露问题成为了一个重要话题。导致用户隐私数据被泄露的原因主要有两种。一种是由于软件编写者的粗心所致,另一种是攻击者所编写的恶意软件的故意行为。但有的开发者为了谋取利润,使用了广告库所开发的非恶意软件同样会泄露用户的隐私数据。研究者通常会根据Android恶意软件的主要威胁特征:违反隐私保护机制,即泄露敏感信息(例如位置信息、联系人信息、图片、短信等)来识别恶意软件。所以,应用中的隐私泄露作为恶意软件的重要特征,引起了学术界的普遍关注。虽然基于Android平台的数据流分析不同于PC端,但是由于学术界众多研究者的共同努力,许多像Activity生命周期,回调函数,组件间通信等问题均己得到了很好的解决。我们在研究基于数据流分析的安卓平台隐私泄露的过程中发现由于Fragment出现的多样性,在回调函数以及Activity生命周期中动态绑定的Fragment占据了很大一部分,而在执行数据流分析时,Fragment的生命周期对数据流分析结果有着重要的影响。因此,缺少对这部分Fragment的分析将导致很大一部分的隐私泄露会被丢失。另一方面,攻击者可能会利用Fragment这种隐蔽点来隐藏其恶意行为,从而绕过恶意软件检测工具的检测。因此如果不完全考虑动态注册的Fragment的生命周期,那么由于该部分Fragment所引起的隐私泄露问题将无法被检测到。通过调研我们发现,在目前解决安卓平台隐私泄露问题的研究工作当中,都没有注意到这一问题。因此我们通过实验证明,执行数据流分析时应用中动态绑定的Fragment对检测结果有着至关重要的影响。我们基于学术界公认的静态数据流分析框架FlowDroid实现了分析结果更加全面的隐私泄露检测工具:FragHarden,并通过大量测试集的实验证明了动静态注册的Fragment对隐私泄露检测结果不同程度的影响,验证了FragHarden的有效性。最后我们对FragHarden的执行效率进行了对比评估,其平均性能开销为16.7%。
[Abstract]:With the development of technology, smart phone brings great convenience to people's life.Phones powered by Android smartphones now account for 81% of the mobile market, making Android an important target for attackers to steal user privacy data, according to statistics.Therefore, privacy disclosure caused by smart phone has become an important topic.There are two main reasons for the disclosure of user privacy data.One is due to the carelessness of the software writer, the other is the intentional behavior of the malicious software written by the attacker.But in order to make profits, some developers use the non-malware developed by advertising library to reveal users' privacy data.Researchers usually identify malware according to the main threat characteristics of Android malware: violation of privacy protection mechanism, i.e. disclosure of sensitive information (such as location information, contact information, pictures, short messages, etc.).Therefore, privacy disclosure in applications as an important feature of malicious software has attracted widespread attention in academia.Although the data flow analysis based on Android platform is different from that of PC, many problems such as Activity lifecycle, callback function and inter-component communication have been solved well due to the joint efforts of many researchers in academic circles.In our study of privacy leaks on Android platforms based on data flow analysis, we found that because of the diversity of Fragment, dynamically bound Fragment in callback functions and Activity lifecycle accounts for a large part of it.However, the life cycle of fragment plays an important role in data flow analysis.Therefore, the lack of analysis of this part of the Fragment will lead to a large part of the privacy disclosure will be lost.On the other hand, an attacker may use a hidden point such as Fragment to hide his malicious behavior, thus circumventing the detection of malware detection tools.Therefore, if the life cycle of dynamically registered Fragment is not fully considered, the privacy disclosure caused by this part of Fragment will not be detected.Through research we found that none of the current research efforts to address privacy leaks on Android platforms had noticed the issue.Therefore, we have proved that the dynamic binding of Fragment in data flow analysis is of great importance to the detection results.Based on the static data flow analysis framework (FlowDroid), we have implemented a more comprehensive privacy leak detection tool:: frame Harden.Through the experiments of a large number of test sets, we have proved that the dynamically and statically registered Fragment has a privacy leak detection knot.Of different degrees of influence,The validity of FragHarden is verified.Finally, we compare and evaluate the performance efficiency of FragHarden, and the average performance cost is 16.7.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309;TP316

【相似文献】