Federal Project Grant Award Summary

AIGIS: Securing the Deep Learning Model Supply Chain

The National Science Foundation (NSF), Division of Computer and Network Systems, awarded Purdue University a $410,000 Project Grant (CFDA 47.070 – Computer and Information Science and Engineering) effective October 1, 2026, through September 30, 2030, to develop security methods and tools for validating the trustworthiness of pre-trained artificial intelligence (AI) models distributed through open online repositories. The research addresses three critical vulnerabilities in the machine learning (ML) model supply chain: model spoofing (adversarial uploads using deceptive naming conventions), insecure deserialization processes vulnerable to arbitrary code execution, and model management threats. The project integrates software engineering principles with machine learning techniques to implement anomaly detection schemes for naming conventions and architectural signatures, develop automated least-privilege deserialization mechanisms, and establish safe model handling subsets to prevent security breaches and operational failures in AI-enabled systems.

This collaborative research initiative will strengthen the U.S. research enterprise and economic competitiveness by improving infrastructure security for AI systems relied upon by research, industry, and government stakeholders. Beyond technical deliverables, the project advances workforce development through student training and research opportunities while fostering collaboration among universities, industry partners, and other stakeholders.