基于SVM的Android应用程序安全检测综述

当前基于SVM的Android应用程序安全检测技术主要是通过将SVM算法与动静态分析方法相结合,应用于Android应用程序的漏洞和恶意软件的检测中,而恶意软件的检测又可分为恶意行为的检测和恶意代码的检测。故本文按SVM算法应用到的检测领域分类,分别对其应用于Android应用程序中的恶意行为检测、恶意代码检测和漏洞检测方面的研究进行分析与讨论,并总结了当前该领域中仍然存在的一些问题,给出了SVM算法和其应用于Android安全检测中的改进之处,最后对未来的发展进行了展望。     

基于沙盒的Android恶意软件动态分析方案

智能手机的普及极大地刺激了恶意软件的广泛传播,Android平台因其巨大的市场占有率和开源特性,已成为攻击者首选的攻击目标。针对传统的基于签名的反病毒软件仅能检测已知恶意软件的缺点,文章提出基于沙盒的Android恶意软件动态分析方案,用于有效地分析未知恶意软件的行为。文章通过在虚拟化软件Oracle VM VirtualBox中安装Android x86虚拟机的方式来实现Android沙盒,利用VirtualBox提供的命令行工具来控制Android沙盒。Android应用程序通过调用相应系统API来完成对应的行为,文中方案通过在应用程序包中插入API监视代码的方法监测Android应用程序调用的系统API,并通过脚本程序向Android沙盒发送不同的用户事件流来模拟用户对应用程序的真实操作,控制Android应用程序在沙盒中自动运行,实验证明文中提出的方法切实可行。     

基于签名与数据流模式挖掘的Android恶意软件检测系统

随着Android软件开发和维护的不断增多,以及恶意软件的抗检测能力逐渐增强,主流的静态检测方法开始面临一些问题:签名检测虽然检测速度快,但是对代码混淆、重打包类的恶意软件的检测能力不强;基于数据流的检测方法虽然精度高,但检测效率低。针对上述技术存在的缺点,提出了一种混合型静态检测系统。该系统改进了多级签名检测方法,通过对method与class签名进行多级匹配,提高了对代码混淆类恶意软件的检测能力。系统还改进了传统数据流分析技术,通过数据流模式挖掘,找出恶意软件频繁使用的数据流模式,省去了人工确认环节,提高了数据流分析的自动化程度与效率。两种技术的结合使得系统在检测精度与效率两方面达到一个合理的折中点。实验结果表明,该系统对于代码混淆和重打包的恶意软件具有较好的检测能力,对主流恶意软件的检测精确度达到88%。     

基于ADB调试的安卓恶意软件研究

目前,安卓手机市场占有率已超过80%,由于安卓系统的开源性和可移植性,导致安卓恶意软件数量呈几何式增长,因此研究简便可行的反制措施十分重要。ADB调试原用于调试应用程序,但对于现阶段恶意软件的提取、分析有巨大帮助。通过研究安卓恶意软件的基本原理,实现安卓恶意软件的检测、分析、应对以及反制,并提出基于ADB调试的安卓恶意软件反制方法。经检测,此方法可以高效分析并反制恶意软件,具有实际应用价值。     

基于可解释性的Android恶意软件检测

针对Android恶意软件检测,通常仅有检测结果缺乏对其检测结果的可解释性.基于此,从可解释性的角度分析Android恶意软件检测,综合利用多层感知机和注意力机制提出一种可解释性的Android恶意软件检测方法(multilayer perceptron attention-method, MLP＿At).通过提取Android恶意软件的应用权限和应用程序接口(application programming interface, API)特征来进行数据预处理生成特征信息,采用多层感知机对特征学习.最后,利用BP算法对学习到的数据进行分类识别.在多层感知机中引入注意力机制,以捕获敏感特征,根据敏感特征生成描述来解释应用的核心恶意行为.实验结果表明所提方法能有效检测恶意软件,与SVM、RF、XGBoost相比准确率分别提高了3.65%、3.70%和2.93%,并能准确地揭示软件的恶意行为.此外,该方法还可以解释样本被错误分类的原因.     

一种基于多特征集成学习的恶意代码静态检测框架

伴随着互联网的普及和5G通信技术的快速发展,网络空间所面临的威胁日益增大,尤其是恶意软件的数量呈指数型上升,其所属家族的变种爆发式增加.传统的基于人工签名的恶意软件的检测方式速度太慢,难以处理每天数百万计新增的恶意软件,而普通的机器学习分类器的误报率和漏检率又明显过高.同时恶意软件的加壳、混淆等对抗技术对该情况造成了更大的困扰.基于此,提出一种基于多特征集成学习的恶意软件静态检测框架.通过提取恶意软件的非PE(Portable Executable)结构特征、可见字符串与汇编码序列特征、PE结构特征以及函数调用关系5部分特征,构建与各部分特征相匹配的模型,采用Bagging集成和Stacking集成算法,提升模型的稳定性,降低过拟合的风险.然后采取权重策略投票算法对5部分集成模型的输出结果做进一步聚合.经过测试,多特征多模型聚合的检测准确率可达96.99%,该结果表明:与其他静态检测方法相比,该方法具有更好的恶意软件鉴别能力,对加壳、混淆等恶意软件同样具备较高的识别率.     

基于非用户操作序列的恶意软件检测方法

针对Android恶意软件持续大幅增加的现状以及恶意软件检测能力不足这一问题,提出了一种基于非用户操作序列的静态检测方法。首先,通过对恶意软件进行逆向工程分析,提取出恶意软件的应用程序编程接口（API）调用信息;然后,采用广度优先遍历算法构建恶意软件的函数调用流程图;进而,从函数流程图中提取出其中的非用户操作序列形成恶意行为库;最后,采用编辑距离算法计算待检测样本与恶意行为库中的非用户操作序列的相似度进行恶意软件识别。在对360个恶意样本和300的正常样本进行的检测中,所提方法可达到90.8%的召回率和90.3%的正确率。与Android恶意软件检测系统Androguard相比,所提方法在恶意样本检测中召回率提高了30个百分点;与FlowDroid方法相比,所提方法在正常样本检测中准确率提高了11个百分点,在恶意样本检测中召回率提高了4.4个百分点。实验结果表明,所提方法提高了恶意软件检测的召回率,有效提升恶意软件的检测效果。     

基于Dalvik指令的Android恶意代码特征描述及验证

为实现Android平台下恶意软件的高效检测,提出了一种基于Dalvik指令的Android恶意代码特征形式化描述和分析方法,能够在无需反编译应用程序的基础上,快速检测样本的恶意特征.该方法首先依照DEX文件格式对Android应用程序切分得到以方法为单位的指令块,通过对块中Dalvik指令进行形式化描述以实现程序特征的简化和提取,之后综合使用改进的软件相似度度量算法和闵可夫斯基距离算法计算提取特征与已知恶意特征的相似度,并根据相似度比对结果来判定当前待测软件是否含有恶意代码.最后建立原型系统模型来验证上述方法,以大量随机样本进行特征匹配实验.实验结果表明,该方法描述特征准确、检测速度较快,适用于Android恶意代码的快速检测.     

一种基于签名和属性的可执行文件比较

可执行文件比较广泛应用于软件版权检测、恶意软件家族检测、异常检测的模式更新以及补丁分析.传统方法无法满足应用对速度和精度的要求.在函数、基本块和指令级别上设计了一元指令签名、基于函数控制流程图邻接矩阵的函数一元结构签名、指令的强/中/弱一元签名,并提出了融合签名和属性的函数匹配算法、基本块匹配算法,从而简化了已有指令比较,可抗指令重排,优于SPP.并通过匹配权统计以及严格的最大唯一匹配策略和Hash进一步降低误报,提高效率.最后,实现原型工具PEDiff,并通过实验证实了该比较方法在速度和精度上具有良好的性能.     

基于权限和API特征结合的Android恶意软件检测方法

随着Android操作系统的广泛应用,基于Android平台的应用程序的数量日益增长。如何有效地识别恶意软件,对保护手机的安全性至关重要。提出了基于权限和API特征结合的Android恶意软件检测方法,该方法通过反编译apk文件来提取权限特征和API特征,并将两者相结合作为一个整体的特征集合。在此基础上,采用分类算法进行恶意软件的甄别。实验结果表明,该方法的判别准确率高于权限集合或API集合单独作为特征的判别方法,从而能更加有效地检测Android恶意应用程序。     

基于多级签名匹配算法的Android恶意应用检测*

针对Android恶意应用泛滥的问题,提出了一种基于恶意应用样本库的多级签名匹配算法来进行Android恶意应用的检测。以MD5哈希算法与反编译生成的smali文件为基础,生成API签名、Method签名、Class签名、APK签名。利用生成的签名信息,从每一类恶意应用样本库中提取出这类恶意行为的共有签名,通过匹配待检测应用的Class签名与已知恶意应用样本库的签名,将待测应用中含有与恶意签名的列为可疑应用,并回溯定位其恶意代码,确定其是否含有恶意行为。在测试中成功的发现可疑应用并定位了恶意代码,证明了本系统的有效性。     

Android application behavioural analysis for data leakage

An android application requires specific permissions from the user to access the system resources and perform required functionalities. Recently, the android market has experienced exponential growth, which leads to malware applications. These applications are purposefully developed by hackers to access private data of the users and adversely affect the application usability. A suitable tool to detect malware is urgently needed, as malware may harm the user. As both malware and clean applications require similar types of permissions, so it becomes a very challenging task to differentiate between them. A novel algorithm is proposed to identify the malware‐based applications by probing the permission patterns. The proposed method uses the k‐means algorithm to quarantine the malware application by obtaining permission clusters. An efficiency of 90% (approx.) is attained for malicious behaviour, which validates this work. This work substantiates the use of application permissions for potential applications in android malware detection.     

Android恶意软件的多特征协作决策检测方法

由于智能手机使用率持续上升促使移动恶意软件在规模和复杂性方面发展更加迅速。作为免费和开源的系统,目前Android已经超越其他移动平台成为最流行的操作系统,使得针对Android平台的恶意软件数量也显著增加。针对Android平台应用软件安全问题,提出了一种基于多特征协作决策的Android恶意软件检测方法,该方法主要通过对Android 应用程序进行分析、提取特征属性以及根据机器学习模型和分类算法判断其是否为恶意软件。通过实验表明,使用该方法对Android应用软件数据集进行分类后,相比其他分类器或算法分类的结果,其各项评估指标均大幅提高。因此,提出的基于多特征协作决策的方式来对Android恶意软件进行检测的方法可以有效地用于对未知应用的恶意性进行检测,避免恶意应用对用户所造成的损害等。     

Mining permission patterns for contrasting clean and malicious android applications

An Android application uses a permission system to regulate the access to system resources and users’ privacy-relevant information. Existing works have demonstrated several techniques to study the required permissions declared by the developers, but little attention has been paid towards used permissions. Besides, no specific permission combination is identified to be effective for malware detection. To fill these gaps, we have proposed a novel pattern mining algorithm to identify a set of contrast permission patterns that aim to detect the difference between clean and malicious applications. A benchmark malware dataset and a dataset of 1227 clean applications has been collected by us to evaluate the performance of the proposed algorithm. Valuable findings are obtained by analyzing the returned contrast permission patterns.     

“Andromaly”: a behavioral malware detection framework for android devices

This article presents Andromaly—a framework for detecting malware on Android mobile devices. The proposed framework realizes a Host-based Malware Detection System that continuously monitors various features and events obtained from the mobile device and then applies Machine Learning anomaly detectors to classify the collected data as normal (benign) or abnormal (malicious). Since no malicious applications are yet available for Android, we developed four malicious applications, and evaluated Andromaly’s ability to detect new malware based on samples of known malware. We evaluated several combinations of anomaly detection algorithms, feature selection method and the number of top features in order to find the combination that yields the best performance in detecting new malware on Android. Empirical results suggest that the proposed framework is effective in detecting malware on mobile devices in general and on Android in particular.     

基于机器学习算法的Android恶意程序检测系统

对于传统的恶意程序检测方法存在的缺点,针对将数据挖掘和机器学习算法被应用在未知恶意程序的检测方法进行研究。当前使用单一特征的机器学习算法无法充分发挥其数据处理能力,检测效果不佳。文中将语音识别模型与随机森林算法相结合,首次提出了综和APK文件多类特征统一建立N-gram模型,并应用随机森林算法用于未知恶意程序检测。首先,采用多种方式提取可以反映Android恶意程序行为的3类特征,包括敏感权限、DVM函数调用序列以及OpCodes特征;然后,针对每类特征建立N-gram模型,每个模型可以独立评判恶意程序行为;最后,3类特征模型统一加入随机森林算法进行学习,从而对Android程序进行检测。基于该方法实现了Android恶意程序检测系统,并对811个非恶意程序及826个恶意程序进行检测,准确率较高。综合各个评价指标,与其他相关工作对比,实验结果表明该系统在恶意程序检测准确率和有效性上表现更优。     

一种基于组合事件行为触发的Android恶意行为检测方法

当前Android恶意应用程序在传播环节缺乏有效的识别手段,对此提出了一种基于自动化测试技术和动态分析技术的Android恶意行为检测方法。 通过自动化测试技术触发Android应用程序的行为,同时构建虚拟的沙箱监控这些行为。设计了一种组合事件行为触发模型——DroidRunner,提高了Android应用程序的代码覆盖率、恶意行为的触发率以及Android恶意应用的检测率。经过实际部署测试,该方法对未知恶意应用具有较高的检测率,能帮助用户发现和分析未知恶意应用。     

Security implications of running windows software on a Linux system using Wine: a malware analysis study

Linux is considered to be less prone to malware compared to other operating systems, and as a result Linux users rarely run anti-malware. However, many popular software applications released on other platforms cannot run natively on Linux. Wine is a popular compatibility layer for running Windows programs on Linux. The level of security risk that Wine poses to Linux users is largely undocumented. This project was conducted to assess the security implications of using Wine, and to determine if any specific types of malware or malware behavior have a significant effect on the malware being successful in Wine. Dynamic analysis (both automated and manual) was applied to 30 malware samples both in a Windows environment and Linux environment running Wine. Behavior analyzed included file system, registry, and network access, and the spawning of processes, and services. The behavior was compared to determine malware success in Wine. The study results provide evidence that Wine can pose serious security implications when used to run Windows software in a Linux environment. Five samples of Windows malware were run successfully through Wine on a Linux system. No significant relationships were discovered between the success of the malware and its high-level behavior or malware type. However, certain API calls could not be recreated in a Linux environment, and led to failure of malware to execute via Wine. This suggests that particular malware samples that utilize these API calls will never run completely successfully in a Linux environment. As a consequence, the success of some samples can be determined from observing the API calls when run within a Windows environment.

     

A new machine learning-based method for android malware detection on imbalanced dataset

Nowadays, malware applications are dangerous threats to Android devices, users, developers, and application stores. Researchers are trying to discover new methods for malware detection because the complexity of malwares, their continuous changes, and damages caused by their attacks have increased. One of the most important challenges in detecting malware is to have a balanced dataset. In this paper, a detection method is proposed to identify malware to improve accuracy and reduce error rates by preprocessing the used dataset and balancing it. To attain these purposes, the static analysis is used to extract features of the applications. The ranking methods of features are used to preprocess the feature set and the low-effective features are removed. The proposed method also balances the dataset by using the techniques of undersampling, the Synthetic Minority Oversampling Technique (SMOTE), and a combination of both methods, which have not yet been studied among detection methods. Then, the classifiers of K-Nearest Neighbor (KNN), Support Vector Machine, and Iterative Dichotomiser 3 are used to create the detection model. The performance of KNN with SMOTE is better than the performance of the other classifiers. The obtained results indicate that the criteria of precision, recall, accuracy, F-measure, and Matthews Correlation Coefficient are over 97%. The proposed method is effective in detecting 99.49% of the malware’s existing in the used dataset and new malware.

     

Android malware detection based on system call sequences and LSTM

As Android-based mobile devices become increasingly popular, malware detection on Android is very crucial nowadays. In this paper, a novel detection method based on deep learning is proposed to distinguish malware from trusted applications. Considering there is some semantic information in system call sequences as the natural language, we treat one system call sequence as a sentence in the language and construct a classifier based on the Long Short-Term Memory (LSTM) language model. In the classifier, at first two LSTM models are trained respectively by the system call sequences from malware and those from benign applications. Then according to these models, two similarity scores are computed. Finally, the classifier determines whether the application under analysis is malicious or trusted by the greater score. Thorough experiments show that our approach can achieve high efficiency and reach high recall of 96.6% with low false positive rate of 9.3%, which is better than the other methods.