基于深度学习的移动应用安全静动态检测方法研究

Research on Static and Dynamic Detection Method of Mobile Application Security Based on Deep Learning

许多Android、iOS、鸿蒙移动操作系统应用依赖的第三方库和SDK会频繁更新,新的版本修复了旧版本中的安全漏洞,但也会引入新的漏洞。各平台系统涉及的应用具有复杂性与动态性,其代码库较为庞大,难以覆盖各平台风险检测点。为此,提出基于深度学习的移动应用安全静动态检测方法研究。采用加权平均滤波和高斯滤波结合的方法,对移动应用流量数据滤波处理,并将其输入至深度学习TensorFlow框架中,提取流量波动幅值特征,计算获取的流量波动幅值特征与静动态检测标准特征之间的匹配度,实现移动应用运行安全静动态检测。实验结果表明,该方法应用下召回率更高,能够检测到更多的安全漏洞,在不同数据量下的响应时间较少。

The third-party libraries and SDKS that many Android,iOS,and Android mobile op-erating system applications rely on are frequently updated,and new versions fix security vulner-abilities in older versions,but also introduce new ones.The application involved in each plat-form system is complex and dynamic,and its code base is rclatively large,which is difficult to cover the risk detection points of each platform.Therefore,the static and dynamic detection method of mobile application security based on deep learning is proposed.The combined method of weighted average filtering and Gaussian filtering is used to filter the traffic data of mobile applications and input it into the deep learning TensorFlow framework to extract the amplitude characteristics of traffic fluctuations and calculate the matching degree between the obtained am-plitude characteristics of traffic fluctuations and the static and dynamic detection standard charac-teristics,so as to realize the static and dynamic detection of the safe operation of mobile applica-tions.The experimental results show that the method has higher recall rate,more security vul-nerabilities can be detected,and less response time under different data volumes.