Ip-Cores Watermarking Scheme At Behavioral Level Using Genetic Algorithms

This paper presents an improved watermarking scheme for soft Intellectual Property (IP)-Cores using Genetic Algorithms (GAs). For this purpose, a watermark signature and an IP-Core behavioral description are translated into Finite State Machines (FSMs). Both FSMs are merged in a single one containing a watermarked IPCore without disrupting its original functionality. Therefore, we tackle NP-completeness of the subgraph isomorphism problem found during FSMs merging process via a tailored GA. However, not deeply embedded states are a key problem that may ease watermark's removal. To overcome this issue, a FSM reduction algorithm is also considered in the proposed watermarking scheme. Moreover, the proposed scheme also targets transitions regrouping which impact the amount of hardware resources usage and a mechanism for selecting the best watermarked FSM. A thorough empirical assessment shows a significant improvement in terms of reduction and watermark embedding strength.