EnsembleDroid: A Malware Detection Approach for Android System based on Ensemble Learning
EnsembleDroid: A Malware Detection Approach for Android System based on Ensemble Learning
复制标题
DOI：
10.1109/urtc56832.2022.10002213
复制
发表时间：
2022-09
期刊：
影响因子：
--
通讯作者：
Sharon Guan;Wei Li
中科院分区：
文献类型：
--
作者：
Sharon Guan;Wei Li
近年来，智能手机和平板电脑等移动设备在功能和便利性方面已成为我们日常生活中最受欢迎的数字设备之一。随着Android操作系统的日益普及，移动设备已经成为恶意软件的常见目标，尤其是通过第三方市场。更糟糕的是，混淆和对抗性示例攻击的出现使恶意软件能够逃避传统的安全方法并窃取用户的私人信息。在本文中，我们提出了一个基于集成学习的框架，通过使用风险权限作为特征来训练分类器并确定移动应用程序（a.k.a. app）是否为恶意软件来检测恶意软件。为了评估所提出的恶意软件检测方法的性能，我们使用由恶意和良性应用组成的真实Android应用数据集进行了一系列实验。实验结果清楚地表明，所提出的恶意软件检测方法能够有效地检测出恶意软件，具有较高的准确率。
In recent years, mobile devices such as smartphones and tablets have become one of the most popular digital devices of choice in our daily lives when it comes to functionality and convenience. With only ever increasing popularity, mobile devices with Android operating systems have become a common target for malware especially through third-party markets. To make things worse, the emergence of obfuscation and adversarial example attacks enables malware to evade traditional security methods and steal a user’s private information. In this paper, we propose an ensemble learning-based framework for detecting malware by using risky permissions as features to train a classifier and determine whether a mobile application (a.k.a. app) is malware or not. To evaluate the performance of the proposed malware detection approach, we have conducted a series of experiments using real world Android app datasets that are composed of both malicious and benign apps. Experimental results clearly show that the proposed malware detection approach can effectively detect malware with a high accuracy.