How Can We Craft Large-Scale Android Malware? An Automated Poisoning Attack
2019 IEEE 1st International Workshop on Artificial Intelligence for Mobile (AI4Mobile)(2019)
摘要
Android malware, is one of the most serious threats to mobile security. Today, machine learning-based approach is one of the most promising approaches in detecting Android malware. However, our previous experiments show that sophisticated attackers can craft large-scale Android malware to pollute training data and pose an automated poisoning attack on machine learning-based malware detection systems (e.g., Drebin, Droidapiminer, Stormdroid, and Mamadroid), and eventually mislead the detection tools. We further examine how machine learning classifiers can be mislead under four different attack models and significantly reduce detection accuracy. Apart from Android malware, to better protect mobile devices, we also discuss a general threat model of Android devices to investigate the capabilities of different attackers.
更多查看译文
关键词
Malware,Mobile handsets,Training,Browsers,Biological system modeling,Machine learning,Mobile applications
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络