Proj CMI Paper Reading: Mining Android Apps for Anomalies
Abstract
本文:
任务:探究根据描述主题对 Android 应⽤程序进行集群,并识别每个集群中的API 异常值
方法:改进CHABADA
实验: 22,500 +个 Android 应⽤程序
效果:
- predict 74% of novel malware as such, without requiring any known malware patterns, maintaining the false positive rate close to 10%