| 摘要: |
| 随着对Android恶意软件检测精度和性能要求的提高，越来越多的Android恶意软件检测引擎使用人工智能算法。与此同时，攻击者开始尝试对Android恶意软件进行一定的修改，使得Android恶意软件可以在保留本身的功能的前提下绕过这些基于人工智能算法的检测。上述过程即是Android恶意软件检测领域的对抗攻击。本文梳理了目前存在的基于人工智能算法的Android恶意软件检测模型，概述了针对Android恶意软件检测模型的对抗攻击方法，并从特征和算法两方面总结了相应的增强模型安全性的防护手段，最后提出了Android恶意软件检测模型和对抗攻击的发展趋势，并分析了对抗攻击对Android恶意软件检测的影响。 |
| 关键词: Android安全 恶意软件 对抗攻击 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2021.07.02 |
| 投稿时间：2019-07-27修订日期：2019-10-08 |
| 基金项目:本课题得到国家重点研发计划（No.2019YFB1706000）资助。 |
|
| A Survey of Adversarial Attack Techniques for Android Malware Detection |
| LI Jialin,WANG Yazhe,LUO Lvgen,WANG Yu |
| Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China |
| Abstract: |
| With the accuracy and performance demand for Android malware detection, more and more Android malware detection engines integrate artificial intelligence algorithms. At the same time, attackers have begun to try to modify Android malware to bypass these artificial intelligence based algorithm while preserving their own functionality. It's called adversarial attack in the field of Android malware detection. This paper combs the existing Android malware detection model based on artificial intelligence algorithm, and summarizes the adversarial attack methods for Android malware detection model and the corresponding protection methods for enhancing model security from two aspects of features and algorithms. Finally, the development trend of Android malware detection model and confrontation attack is proposed, and the impact of adversarial attack on Android malware detection is analyzed. |
| Key words: android security malware detection adversarial attack |
