AI and ML Attacks on IC Hardware Security - Demonstration for Cybersecurity Students
Abstract
The IC design and security industry depends on trusted systems, yet remains challenged by an increasingly fragmented supply chain and evolving threat landscape. The rise of fabless enterprises and the proliferation of AI/ML technologies have further exposed hardware security to new vulnerabilities. This paper provides proof-of-concept implementations of emerging threats posed by machine learning to IC hardware design, focusing on two distinct areas: GNN-based attacks on logic locking and the insertion of hardware Trojans via large language models. These represent growing and independent research directions in hardware security. We showcase and analyze two representative examples from each category to highlight the risks of unmitigated ML-driven attacks.
AI and ML Attacks on IC Hardware Security - Demonstration for Cybersecurity Students
The IC design and security industry depends on trusted systems, yet remains challenged by an increasingly fragmented supply chain and evolving threat landscape. The rise of fabless enterprises and the proliferation of AI/ML technologies have further exposed hardware security to new vulnerabilities. This paper provides proof-of-concept implementations of emerging threats posed by machine learning to IC hardware design, focusing on two distinct areas: GNN-based attacks on logic locking and the insertion of hardware Trojans via large language models. These represent growing and independent research directions in hardware security. We showcase and analyze two representative examples from each category to highlight the risks of unmitigated ML-driven attacks.