Standardizing Hardware Security Research: Attacks, Defenses, and Benchmarks

• Nicholas West, Computer Engineering, University of Delaware

• Satwik Patnaik, Electrical and Computer Engineering, University of Delaware

As demand for low-power and high-performance electronic devices grows, the fabrication of integrated circuit (IC) chips has become increasingly complex. To manage the chip design complexity, companies outsource chip manufacturing to overseas foundries, raising serious concerns about IP theft, misuse, and tampering during fabrication. Hardware obfuscation has become a crucial defense, intentionally modifying chip designs to make reverse engineering, cloning, and malicious alteration significantly more difficult while still maintaining correct functionality.

However, research progress in hardware obfuscation is hampered by a major challenge: the scarcity of standardized datasets and common benchmarks. Unlike fields such as computer vision, hardware security lacks widely available repositories of diverse obfuscation schemes and associated attacks/defenses. This makes it hard to rigorously evaluate new methods or compare the effectiveness of machine learning approaches for hardware security.

To address this gap, this research introduces a comprehensive, easy-to-use framework that brings together state-of-the-art tools for evaluating both hardware obfuscation (“defenses”) and deobfuscation (“attacks”). By providing a centralized repository of attacks, defenses, and benchmark datasets, our framework accelerates testing and development for both traditional and machine learning-based approaches, ultimately enabling more reliable hardware security research.