$\alpha$ Diff: Cross-Version Binary Code Similarity Detection with DNN

Bingchang Liu, Wei Huo, Chao Zhang, Wenchao Li, Feng Li, Aihua Piao, Wei Zou
{"title":"$\\alpha$ Diff: Cross-Version Binary Code Similarity Detection with DNN","authors":"Bingchang Liu, Wei Huo, Chao Zhang, Wenchao Li, Feng Li, Aihua Piao, Wei Zou","doi":"10.1145/3238147.3238199","DOIUrl":null,"url":null,"abstract":"Binary code similarity detection (BCSD) has many applications, including patch analysis, plagiarism detection, malware detection, and vulnerability search etc. Existing solutions usually perform comparisons over specific syntactic features extracted from binary code, based on expert knowledge. They have either high performance overheads or low detection accuracy. Moreover, few solutions are suitable for detecting similarities between cross-version binaries, which may not only diverge in syntactic structures but also diverge slightly in semantics. In this paper, we propose a solution $\\alpha$ Diff, employing three semantic features, to address the cross-version BCSD challenge. It first extracts the intra-function feature of each binary function using a deep neural network (DNN). The DNN works directly on raw bytes of each function, rather than features (e.g., syntactic structures) provided by experts. $\\alpha$ Diff further analyzes the function call graph of each binary, which are relatively stable in cross-version binaries, and extracts the inter-function and inter-module features. Then, a distance is computed based on these three features and used for BCSD. We have implemented a prototype of $\\alpha$ Diff, and evaluated it on a dataset with about 2.5 million samples. The result shows that $\\alpha$ Diff outperforms state-of-the-art static solutions by over 10 percentages on average in different BCSD settings.","PeriodicalId":6622,"journal":{"name":"2018 33rd IEEE/ACM International Conference on Automated Software Engineering (ASE)","volume":"75 1","pages":"667-678"},"PeriodicalIF":0.0000,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"148","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 33rd IEEE/ACM International Conference on Automated Software Engineering (ASE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3238147.3238199","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 148

Abstract

Binary code similarity detection (BCSD) has many applications, including patch analysis, plagiarism detection, malware detection, and vulnerability search etc. Existing solutions usually perform comparisons over specific syntactic features extracted from binary code, based on expert knowledge. They have either high performance overheads or low detection accuracy. Moreover, few solutions are suitable for detecting similarities between cross-version binaries, which may not only diverge in syntactic structures but also diverge slightly in semantics. In this paper, we propose a solution $\alpha$ Diff, employing three semantic features, to address the cross-version BCSD challenge. It first extracts the intra-function feature of each binary function using a deep neural network (DNN). The DNN works directly on raw bytes of each function, rather than features (e.g., syntactic structures) provided by experts. $\alpha$ Diff further analyzes the function call graph of each binary, which are relatively stable in cross-version binaries, and extracts the inter-function and inter-module features. Then, a distance is computed based on these three features and used for BCSD. We have implemented a prototype of $\alpha$ Diff, and evaluated it on a dataset with about 2.5 million samples. The result shows that $\alpha$ Diff outperforms state-of-the-art static solutions by over 10 percentages on average in different BCSD settings.
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
$\alpha$ Diff:基于DNN的跨版本二进制代码相似度检测
二进制代码相似度检测(BCSD)有许多应用,包括补丁分析、剽窃检测、恶意软件检测和漏洞搜索等。现有的解决方案通常基于专家知识,对从二进制代码中提取的特定语法特征进行比较。它们要么性能开销高,要么检测精度低。此外,很少有解决方案适合检测跨版本二进制文件之间的相似性,这些解决方案不仅在语法结构上存在差异,而且在语义上也有轻微的差异。在本文中,我们提出了一个解决方案$\alpha$ Diff,采用三个语义特征,以解决跨版本BCSD的挑战。首先利用深度神经网络(DNN)提取每个二元函数的函数内特征;DNN直接处理每个函数的原始字节,而不是专家提供的特征(例如语法结构)。$\alpha$ Diff进一步分析每个在跨版本二进制文件中相对稳定的二进制文件的函数调用图,并提取函数间和模块间的特征。然后,根据这三个特征计算距离并用于BCSD。我们已经实现了$\alpha$ Diff的原型,并在大约250万个样本的数据集上对其进行了评估。结果表明,在不同的BCSD设置中,$\alpha$ Diff比最先进的静态解决方案平均高出10个百分点以上。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 去求助
来源期刊
自引率
0.00%
发文量
0
期刊最新文献
Automatically Testing Implementations of Numerical Abstract Domains Self-Protection of Android Systems from Inter-component Communication Attacks Characterizing the Natural Language Descriptions in Software Logging Statements DroidMate-2: A Platform for Android Test Generation CPA-SymExec: Efficient Symbolic Execution in CPAchecker
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1