This $600,000 Project Grant award from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070) aims to advance the security and privacy of on-device machine learning (ML) models used in scientific Internet of Things (IoT)-driven applications.
The project has two primary objectives: 1) Developing a novel runtime detection and prevention mechanism to protect against ML model extraction attacks that could compromise safety- and security-critical models deployed in scientific cyberinfrastructure, and 2) Implementing a comprehensive assessment framework to evaluate the security of on-device ML models. The University of Texas at Arlington is a sub-awardee, managing the scientific and administrative coordination of the project. The goal is to significantly reduce the attack surface for ML models used in critical scientific applications across various domains, enabling more secure deployment of these models in IoT-driven cyberinfrastructure.
Generated 3/25/25, 4:06 AM