EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

LLM-Triaged Vulnerability Analysis for Compiled DNNs

Minghui Li, Yang Li, Changchun Yin, Mingwei Li, Liming Fang

ACM Transactions on Software Engineering and Methodology · 30.11.2026 · 35 (11) · 1-48

DOI: 10.1145/3814615 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACM Transactions on Software Engineering and Methodology, online erschienen am 03.11.2026. Verfasst von Minghui Li und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Compiler toolchains such as TVM are now a standard substrate for deploying deep neural networks, but the optimized executables also expose structures that attackers can leverage. Existing mitigations are commonly applied as a global composite defense, which is effective yet costly because it disrupts performance-critical regularity everywhere. We present VulnSketch, the first LLM-driven vulnerability analysis system for compiled DNNs, and introduce an evidence-bounded formulation of vulnerability analysis that is both auditable and actionable. VulnSketch extracts a Minimal Evidence Interface (MEI) from TVM TIR, scores each operator location along three attacker-aligned vulnerability faces (Segmentation, Regularity, and Lifting), and routes only ambiguous gray locations to bounded LLM triage that must cite MEI fields and propose a pre-approved, semantics-preserving schedule knob. Every proposal is executed under compiler-side legality guards and accepted only if a single micro-validation confirms the expected one-step movement on the prioritized face. Across five representative attacks, VulnSketch attains strong operator-level overlap at small budgets (top 5% recall 0.71–0.80; top 10% recall 0.84–0.91) and high global ranking quality (AUROC 0.85–0.89; AUPRC 0.51–0.62). Applying micro-validated knobs causally reduces attack success by 33–38% and recovered footprint by 38–40%. Operationally, VulnSketch yields 5.9 seconds amortized per verified note under default gray exposure.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Li, M., Li, Y., Yin, C., Li, M., & Fang, L. (2026). LLM-Triaged Vulnerability Analysis for Compiled DNNs. ACM Transactions on Software Engineering and Methodology, 35(11), 1-48. https://doi.org/10.1145/3814615

Vancouver

Li M, Li Y, Yin C, Li M, Fang L. LLM-Triaged Vulnerability Analysis for Compiled DNNs. ACM Transactions on Software Engineering and Methodology. 2026;35(11):1-48. doi:10.1145/3814615

BibTeX
@article{Li2026,
  author = {Li, Minghui and Li, Yang and Yin, Changchun and Li, Mingwei and Fang, Liming},
  title = {LLM-Triaged Vulnerability Analysis for Compiled DNNs},
  journal = {ACM Transactions on Software Engineering and Methodology},
  year = {2026},
  volume = {35},
  number = {11},
  pages = {1--48},
  doi = {10.1145/3814615}
}
Kostenlos starten →