This $207,962 National Science Foundation (NSF) Computer and Information Science and Engineering (CFDA 47.070) grant awarded to Tufts University on February 1, 2024 will support research to enhance the robustness of artificial intelligence (AI) systems against hardware-oriented vulnerabilities. The project has four main thrusts:
-
Designing algorithm-hardware collaborative backdoor attacks to exploit new adversarial vulnerabilities in deep neural networks.
-
Developing methodologies to incorporate the hardware aspect into defenses for improving robustness against supply chain attacks.
-
Creating novel signature embedding frameworks to protect the integrity of deep neural network models in untrusted supply chains.
-
Modeling recovery strategies to mitigate hardware-oriented fault attacks in untrusted environments.
This work aims to ensure trust in AI systems from both the algorithm and hardware perspectives, enabling secure deployment in applications like healthcare, autonomous vehicles, and the Internet of Things. The project results will be integrated into educational programs and used to raise public awareness about AI security.
Generated 12/31/24, 11:06 AM