{ "added_at": "2026-10-19", "arxiv_id": "2107.13404", "arxiv_version_reviewed": "v4", "authors": [ "James Patrick-Evans", "Moritz Dannehl", "Johannes Kinder" ], "bibtex_url": "https://awesomerea.org/papers/citations/paper-xfl.bib", "citation": { "container_title": "2023 IEEE Symposium on Security and Privacy (SP)", "pages": "2375-2390", "publisher": "IEEE", "sources": [ "https://ieeexplore.ieee.org/document/10179439", "https://api.crossref.org/works/10.1109/SP46215.2023.10179439" ] }, "code_url": "https://github.com/lmu-plai/xfl", "date_basis": "IEEE Xplore and the publisher’s Crossref deposit both record publication in May 2023; IEEE added the article to Xplore on 21 July 2023. No publication day is asserted.", "date_basis_zh": "IEEE Xplore 与出版方提交的 Crossref 元数据均记录出版时间为 2023 年 5 月；IEEE 于 2023 年 7 月 21 日将文章加入 Xplore。本条目不认定具体出版日。", "date_precision": "month", "description_en": "XFL treats function naming in stripped binaries as extreme multi-label classification: names are split into tokens, and PfastreXML ranks which tokens from a vocabulary of frequent labels fit each function. Its DEXTER embedding combines static-analysis features built on angr’s VEX IR with call-graph and whole-binary context, and a separate language model orders predicted tokens into a name string. On 741,724 functions from 10,047 x86-64 C binaries in Debian packages, with a 1,024-label vocabulary, the authors report precision 0.835, recall 0.575 and F1 0.681, ahead of Punstrip, Debin and Nero in precision and F1.", "description_zh": "XFL 将去符号二进制中的函数命名视为极端多标签分类：函数名被拆分为词元，再由 PfastreXML 从常见标签词表中为每个函数排序合适的词元。其 DEXTER 嵌入结合基于 angr VEX 中间表示的静态分析特征、调用图上下文和整个二进制的全局上下文，另有一个语言模型把预测的词元排列成函数名字符串。在来自 Debian 软件包、共 10,047 个 x86-64 C 二进制中的 741,724 个函数上，以 1,024 个标签的词表，作者报告精确率 0.835、召回率 0.575、F1 0.681，在精确率和 F1 上领先 Punstrip、Debin 与 Nero。", "doi": "10.1109/SP46215.2023.10179439", "entry_url": "https://awesomerea.org/papers/paper-xfl/", "entry_url_zh": "https://awesomerea.org/zh/papers/paper-xfl/", "evidence_urls": [ "https://ieeexplore.ieee.org/document/10179439", "https://api.crossref.org/works/10.1109/SP46215.2023.10179439", "https://arxiv.org/abs/2107.13404", "https://arxiv.org/pdf/2107.13404v4", "https://github.com/lmu-plai/xfl/blob/5ad25691972158d4f559c401a7462f82fc5412a0/README.md", "https://github.com/lmu-plai/xfl/blob/5ad25691972158d4f559c401a7462f82fc5412a0/COPYING", "https://github.com/lmu-plai/xfl/blob/5ad25691972158d4f559c401a7462f82fc5412a0/xfl-r/README.md", "https://github.com/lmu-plai/xfl/blob/5ad25691972158d4f559c401a7462f82fc5412a0/xfl-r/INSTALL.md", "https://github.com/lmu-plai/xfl/blob/5ad25691972158d4f559c401a7462f82fc5412a0/xfl/evaluation/README.md", "https://zenodo.org/records/10733597" ], "first_arxiv_date": "2021-07-28", "id": "paper-xfl", "json_url": "https://awesomerea.org/papers/paper-xfl/record.json", "kind": "research", "markdown_url": "https://awesomerea.org/papers/paper-xfl/index.md", "markdown_url_zh": "https://awesomerea.org/zh/papers/paper-xfl/index.md", "metadata_notes": "IEEE Xplore and Crossref agree on the title, three-author order, DOI, pages 2375–2390 and the May 2023 date. Reviewed arXiv v4 (2022-12-12; v1 2021-07-28), whose title and authors match. The paper’s code link github.com/unibw-patch/xfl now redirects to lmu-plai/xfl. The reviewed 2025-11-30 snapshot carries GPL-3.0 in COPYING. Its README describes xfl/ as partial original source with experiment data and logs and recommends the refactored xfl-r/ for comparisons; lm/ holds the name-generation language model. The README’s BibTeX lists pages 1677–1692 and DOI 10.1109/SP46215.2023.00096, which doi.org does not resolve. XFL-R setup requires Python 3.8, TensorFlow 2.13.1, PostgreSQL 13, radare2 5.5.4 and Ghidra 10.4, a PfastreXML build compiled from source downloaded separately, and a 46.9 GB Zenodo archive of models and tables (10.5281/zenodo.10733597) whose metadata names no license; pretrained DEXTER embeddings are linked from a separate download host. The paper reports about 850 GB of RAM for DEXTER training.", "metadata_notes_zh": "IEEE Xplore 与 Crossref 的标题、三位作者顺序、DOI、页码 2375–2390 及 2023 年 5 月日期一致。审阅 arXiv v4（2022-12-12；v1 为 2021-07-28），其标题与作者一致。论文中的代码链接 github.com/unibw-patch/xfl 现重定向至 lmu-plai/xfl。所审阅的 2025-11-30 快照在 COPYING 中采用 GPL-3.0。其 README 说明 xfl/ 是包含实验数据和日志的部分原始源码，并建议使用重构后的 xfl-r/ 进行比较；lm/ 包含生成函数名的语言模型。README 中的 BibTeX 列出页码 1677–1692 及 DOI 10.1109/SP46215.2023.00096，但 doi.org 无法解析该 DOI。XFL-R 的安装需要 Python 3.8、TensorFlow 2.13.1、PostgreSQL 13、radare2 5.5.4 与 Ghidra 10.4，需要从另行下载的源码编译 PfastreXML，还需要一个 46.9 GB 的 Zenodo 模型与数据表压缩包（10.5281/zenodo.10733597），其元数据未注明许可证；预训练 DEXTER 嵌入链接至另一个下载站点。论文报告训练 DEXTER 约需 850 GB 内存。", "pdf_url": "https://arxiv.org/pdf/2107.13404v4", "project_url": null, "publication_date": "2023-05", "publication_status": "conference-paper", "record_type": "paper", "runtime_tested": false, "scope_limitations_en": "Predictions are token labels from a fixed vocabulary, ranked and optionally assembled into names; they are not recovered source names, and scoring matches tokens rather than exact names. Evaluation uses x86-64 ELF C binaries from Debian packages, built with mixed compilers and optimization settings, plus Nero’s smaller dataset; the trained model is specific to x86-64 GNU/Linux binaries. All tools receive function boundaries from symbol tables at test time, so boundary-detection errors are excluded. Many test names also occur in training; restricted to names never seen in training, F1 falls to 0.242. Obfuscated and adversarial binaries are out of scope. Debin and Dire figures on the Nero dataset are copied from the Nero paper rather than rerun.", "scope_limitations_zh": "预测结果是固定词表中的词元标签，经排序后可选地组合成名称；它们并非恢复的源码名称，评分也按词元而非完整名称匹配。评测使用来自 Debian 软件包、以多种编译器与优化设置构建的 x86-64 ELF C 二进制，以及规模较小的 Nero 数据集；训练所得模型仅适用于 x86-64 GNU/Linux 二进制。测试时所有工具均从符号表获取函数边界，因此不计入边界检测错误。许多测试函数名也出现在训练集中；仅考察训练中从未出现的名称时，F1 降至 0.242。混淆和对抗性二进制不在研究范围内。Nero 数据集上 Debin 与 Dire 的数据取自 Nero 论文，并未重新运行。", "targets": [ "binary" ], "title": "XFL: Naming Functions in Binaries with Extreme Multi-label Learning", "title_zh": "XFL：基于极端多标签学习的二进制函数命名", "url": "https://ieeexplore.ieee.org/document/10179439", "venue": "IEEE S&P 2023", "verification": "source-reviewed", "verified_at": "2026-10-18", "why_fits_en": "Adds function-name prediction for stripped binaries framed as multi-label learning over binary embeddings, complementing variable-name, type and data-structure recovery research.", "why_fits_zh": "补充以二进制嵌入上的多标签学习预测去符号二进制函数名的方法，与变量名、类型及数据结构恢复研究互补。", "year": 2023 }