基于大数据&机器学习的Android病毒软件检测SVM模型研究

Research on SVM Model of Android Virus Software Detection Based on Big Data & Machine Learning

Android作为当今最流行的操作系统,已经被世界上成千上万的使用者所使用,但是正是基于Android天生开源的特性,让很多开发者在开发app的过程中,可以利用系统或者是手机自身的权限申请以及目前市面上还不健全的应用商店审核机制,开发恶意软件从而危害到用户的手机内的个人资料和隐私。在此篇论文中,我们提供了一种能够检测恶意软件的机制并加以实践,通过提取Android APK的权限[1]申请作为特征点,通过已经使用大量提取的数据训练完的SVM模型对软件做检测,根据实验结果,我们的模型对于恶意软件的检测率高达89%-92%之间,符合研究预期。

Android, as the most popular operating system, has been used by thousands of users in the world, but it is based on the natural open source feature of Andoird that many developers can make use of the permission application of the system or mobile phone and the imperfect application store audit mechanism in the market to develop malicious software to endanger the personal data and privacy of users in the process of developing app. In this paper, we provide a mechanism that can detect malicious software and put it into practice. By extracting the permission of Android APK[1]as a feature point, we use the SVM model trained with a large number of extracted data to detect the software. According to the experimental results, the detection rate of our model for malicious software is as high as 89%, which is in line with the research expectations.