# JsDeObsBench: Measuring and Benchmarking LLMs for JavaScript Deobfuscation

> Builds an execution-verifiable JavaScript deobfuscation benchmark by applying seven JavaScript-Obfuscator transformations to Project CodeNet programs with test cases, then scores LLM and tool outputs for syntax, execution, code simplification and similarity to the original code.

- Authors: Guoqiang Chen; Xin Jin; Zhiqiang Lin
- Venue: ACM CCS 2025
- Status: Conference paper
- Publication / posting date: November 19, 2025
- Date source: ACM-deposited Crossref print-publication date; the October 13–17 CCS 2025 conference week is not used as the publication date.
- First arXiv submission: June 25, 2025
- arXiv: 2506.20170
- Reviewed arXiv version: v1
- DOI: 10.1145/3719027.3744871
- Sources reviewed: 2026-10-17
- Added to the catalog: 2026-10-17
- Original paper: <https://doi.org/10.1145/3719027.3744871>
- PDF: <https://arxiv.org/pdf/2506.20170v1>
- Code: <https://github.com/Ch3nYe/JsDeObsBench>
- Project: <https://jsdeobf.github.io/>
- BibTeX: <https://awesomerea.org/papers/citations/paper-jsdeobsbench.bib>
- Catalog page: <https://awesomerea.org/papers/paper-jsdeobsbench/>
- Record JSON: <https://awesomerea.org/papers/paper-jsdeobsbench/record.json>
- Chinese version: [简体中文](https://awesomerea.org/zh/papers/paper-jsdeobsbench/index.md)

## Scope and limitations

On average, LLM outputs passed execution checks 60.93% of the time, against 83% for the two existing deobfuscator baselines. All samples come from a single obfuscator applied to moderate-length competitive-programming solutions. The 4,515 obfuscated malware samples have no test cases and are compared only through behavior traces.

## Why it fits this collection

Adds a measurable, execution-checked benchmark for LLM-assisted JavaScript deobfuscation, a common step in recovering website behavior from shipped scripts.

## Publication and review notes

ACM-deposited Crossref metadata records print publication on 2025-11-19 and online publication on 2025-11-22, pages 36–50. The reviewed text is arXiv v1, which carries the CCS 2025 header. The paper links the code repository and the jsdeobf.github.io leaderboard. At the pinned commit the repository contains the CodeNet-derived obfuscated datasets, evaluators and scripts, but no malware samples or license file.

## Primary sources

1. <https://doi.org/10.1145/3719027.3744871>
2. <https://api.crossref.org/works/10.1145/3719027.3744871>
3. <https://arxiv.org/abs/2506.20170>
4. <https://github.com/Ch3nYe/JsDeObsBench/blob/8b2e1c35dc4ae25a46ea580ea5dcb6b5de5edbe9/README.md>
5. <https://github.com/Ch3nYe/JsDeObsBench/tree/8b2e1c35dc4ae25a46ea580ea5dcb6b5de5edbe9>
6. <https://jsdeobf.github.io/>

## About this entry

Reviewed against the paper and its primary sources. Cite the original authors using the linked paper or BibTeX.
Original editorial content: MIT. Linked projects, papers and assets retain their own rights and licenses.