DroidGAN:基于DCGAN的Android对抗样本生成框架  

DroidGAN: Android adversarial sample generation framework based on DCGAN

作　　者：唐川[1] 张义[1] 杨岳湘[1] 施江勇[1] TANG Chuan;ZHANG Yi;YANG Yuexiang;SHI Jiangyong(College of Computer,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《通信学报》2018年第A01期64-69,共6页Journal on Communications

基　　金：教育部-中国移动科研基金资助项目(No.MCM20170404)~~

摘　　要：为了加强对Android恶意软件检测系统识别对抗样本的能力,提出了基于深度卷积生成对抗网络的Android对抗样本生成框架。此框架模拟了恶意软件制作者的攻击行为,并在提出的ASG算法的帮助下实现对恶意软件的修改。此框架生成的恶意软件可以绕过检测系统的检测,并能够实际运行而不影响其原有恶意功能。生成的对抗样本可用于重训练原始检测系统,提高系统应对对抗样本的能力。In order to enhance the ability of Android malware detection system to identify adversarial samples,an Android adversarial sample generation framework was proposed based on deep convolution generation adversarial network.This framework simulates the attack behavior of malware makers and implements modifications to malware with the help of proposed ASG algorithm.Malware generated by this framework can bypass the detection system and can actually run without affecting its original malicious functions.The generated adversarial samples can be used to retrain the original detection system and improve the system's ability against adversarial samples.

关 键 词：ANDROID 恶意软件检测 对抗样本 生成对抗网络 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]