Metraipolis › archive › dc_4fdb30bece2c
Post-Hoc Attention Steering of Large Language Models for Robust Code Understanding under Obfuscation — arXiv:2609.26102
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.26102
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T15:16:10.650Z
- SHA-256 of the archived text
9d3e4ccb10c6c62ae3da0b3949addc55fde97293e6b650f81b873848c224522e- Stable URL of this page
https://metraipolis.com/documento/dc_4fdb30bece2c
How to cite
Post-Hoc Attention Steering of Large Language Models for Robust Code Understanding under Obfuscation — arXiv:2609.26102 — arXiv (en), https://arxiv.org/abs/2609.26102 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4fdb30bece2c (Metraipolis archive dc_4fdb30bece2c, sha256 9d3e4ccb10c6c62ae3da0b3949addc55fde97293e6b650f81b873848c224522e)
BibTeX
@misc{dc4fdb30bece2c,
title = {Post-Hoc Attention Steering of Large Language Models for Robust Code Understanding under Obfuscation — arXiv:2609.26102},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.26102},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4fdb30bece2c; sha256 9d3e4ccb10c6c62ae3da0b3949addc55fde97293e6b650f81b873848c224522e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26102. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Code obfuscation is widely used in software systems and malware to conceal program logic and hinder analysis, posing significant challenges for both human developers and automated tools. While large language models (LLMs) have shown strong capabilities in code understanding, their robustness to obfuscation remains poorly understood. Our preliminary study shows that LLM performance significantly degrades on obfuscated code, suggesting a reliance on superficial lexical cues rather than deep semantic reasoning. To address this limitation, we propose CodeSteer, a novel attention steering approach that reallocates model attention toward semantically relevant program elements, including backward slices for output prediction and control-flow paths for execution reasoning. Our method integrates lightweight program analysis with inference-time attention steering to guide LLMs toward the core input-to-output dependencies of a program. Experiments across multiple models and datasets demonstrate that CodeSteer significantly improves performance on obfuscated code, often recovering comparable accuracy to the level of unobfuscated programs. We also show CodeSteer's practical utility through a case study on buffer overflow detection, highlighting its potential for malware/vulnerability analysis and reverse engineering of obfuscated code. Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-23 in cs.SE). Categorie arXiv: Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2609.26102 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T14:42:48.176Z dalla pagina https://arxiv.org/catchup/cs.SE/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).