VexIR2Vec: An Architecture-Neutral Embedding Framework for Binary Similarity
Binary similarity helps detect vulnerabilities, analyze malware, and identify copyright violations, but variations in compiler settings, architectures, and obfuscations complicate the task. We propose VexIR2Vec, an architecture-neutral approach leveraging VEX-IR to address these challenges. VexIR2Vec extracts structural features via peepholes, normalizes them using compiler-inspired transformations (VexINE), and embeds them using a feed-forward Siamese network (VexNet) for binary diffing and searching.
Evaluated against five baselines on a dataset of 2.7M functions and 15.5K binaries across x86 and ARM, VexIR2Vec outperforms the nearest competitor in cross-optimization, cross-compilation, cross-architecture, and obfuscation settings by 40%, 18%, 21%, and 60%, respectively. It also improves search precision by 46%, achieving a mean average precision of 0.76. Built as a lightweight, multithreaded library, VexIR2Vec is ≈ 3.1–3.5× faster than the closest baselines and is significantly scalable.
Pointer to the Journal Version - https://dl.acm.org/doi/10.1145/3721481
Mon 23 JunDisplayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change
10:30 - 12:30 | |||
10:30 20mTalk | An empirical study of business process models and model clones on GitHub Journal First Mahdi Saeedi Nikoo Eindhoven University of Technology, Sangeeth Kochanthara Netherlands' Space Obervatory - ASTRON, Önder Babur Eindhoven University of Technology, Mark van den Brand Eindhoven University of Technology | ||
10:50 20mTalk | The Struggles of LLMs in Cross-lingual Code Clone Detection Research Papers Micheline Bénédicte MOUMOULA University of Luxembourg, Abdoul Kader Kaboré University of Luxembourg, Jacques Klein University of Luxembourg, Tegawendé F. Bissyandé University of Luxembourg DOI | ||
11:10 20mTalk | Clone Detection for Smart Contracts: How Far Are We? Research Papers Zuobin Wang Zhejiang University, Zhiyuan Wan Zhejiang University, Yujing Chen Zhejiang University, Yun Zhang Hangzhou City University, David Lo Singapore Management University, Difan Xie Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security, Xiaohu Yang Zhejiang University DOI | ||
11:30 20mTalk | Measuring Model Alignment for Code Clone Detection Using Causal Interpretation Journal First Shamsa Abid National University of Computer and Emerging Sciences, Xuemeng Cai Singapore Management University, Lingxiao Jiang Singapore Management University | ||
11:50 20mTalk | An Empirical Study of Code Clones from Commercial AI Code Generators Research Papers Weibin Wu Sun Yat-sen University, Haoxuan Hu Sun Yat-sen University, China, Zhaoji Fan Sun Yat-sen University, Yitong Qiao Sun Yat-sen University, China, Yizhan Huang The Chinese University of Hong Kong, Yichen LI The Chinese University of Hong Kong, Zibin Zheng Sun Yat-sen University, Michael Lyu Chinese University of Hong Kong DOI | ||
12:10 20mTalk | VexIR2Vec: An Architecture-Neutral Embedding Framework for Binary Similarity Journal First S. VenkataKeerthy IIT Hyderabad, Soumya Banerjee IIT Hyderabad, Sayan Dey IIT Hyderabad, Yashas Andaluri IIT Hyderabad, Raghul PS IIT Hyderabad, Subrahmanyam Kalyanasundaram IIT Hyderabad, Fernando Magno Quintão Pereira Federal University of Minas Gerais, Ramakrishna Upadrasta IIT Hyderabad | ||
Aurora A is the first room in the Aurora wing.
When facing the main Cosmos Hall, access to the Aurora wing is on the right, close to the side entrance of the hotel.