Hunting Vulnerability Variants in AI Infra: Measurement and Reference-Driven Detection
For AI infrastructure developers and security researchers, this work provides a method to automatically detect cross-repository vulnerability variants, addressing a poorly understood security risk.
The paper measures the prevalence of vulnerability variants in AI infrastructure, finding that 688 repositories and 251 disclosed vulnerabilities show frequent recurrent patterns. It proposes INFRASCOPE, a reference-driven multi-agent framework that detects such variants, uncovering over 20 vulnerabilities including 11 acknowledged and 4 CVEs.
AI infra has become a shared execution layer for model training, deployment, and agent orchestration. Because many projects reimplement similar model-centric workflows, a vulnerability disclosed in one repository can recur as a variant in another repository with a related design. Yet the prevalence and detectability of these variants remain poorly understood. This paper presents a measurement study of vulnerability variants in AI infra. Analyzing 688 GitHub repositories and 251 publicly disclosed vulnerabilities, we find that AI infra projects frequently share overlapping functionality and recurrent vulnerable patterns, creating a concrete basis for cross-repository variants. Building on this finding, we study how to automatically identify such variants from known disclosures. We propose INFRASCOPE, a reference-driven multi-agent framework that extracts transferable vulnerability semantics from known cases and uses them to locate and validate variants in new repositories. Evaluating INFRASCOPE on 20 real-world AI infra repositories, we uncover over 20 vulnerabilities, including 11 acknowledged cases and 4 cases that have been assigned CVEs so far.