Pattern Based Synthetic Benchmark Generation for Hardware Security Applications

The limitations of currently available benchmarks, characterized by their small size, lack of scalability, and inflexibility, present significant challenges in comprehensively evaluating applications in the field of hardware security. This paper aims to illustrate the effective adaptation of attributed graph transformation systems for the automated generation of synthetic benchmark circuits that satisfy constraints on diverse structural features. Additionally, it highlights the flexibility and scalability of this method through the utilization of existing benchmarks by extracting and utilizing frequently occurring subcircuit patterns. A comprehensive description of the methodology is provided, accompanied by a demonstration of the practicality of the generated circuits in hardware security applications, specifically Trojan insertion and detection.