Fingerprinting for IP USER’S Right Protection
Kluwer Academic Publishers eBooks · 2006
In this chapter, we discuss another part of the intellectual property protection, namely how to protect the right of legal IP buyers. In particular, we provide the symmetric fingerprinting techniques such that both IP provider’s illegal distribution and IP buyers’ collusion is discouraged. A fingerprinted IP will not directly prevent misuse of the IP, but will allow the IP provider to detect the source of the redistributed IP and therefore trace the traitor. Fingerprinting-based IP has major advantages over watermarking-based intellectual property protection because it provides protection to both the buyer and seller. The key problem related to the use of fingerprinting for intellectual property protection is the tradeoff between collusion resiliency and runtime. Previous fingerprinting IP protection technique is applicable only to a very restricted set of problems[97]. We have introduced two generic fingerprinting technique for IP protection of solutions to optimization/decision problems and, therefore, of hardware and software intellectual property. By judiciously exploiting partial solution reuse and the incremental application of iterative optimizers, our first set of fingerprinting-based IP protection techniques for partitioning, graph coloring, satisfiability and placement, simultaneously provide high collusion resiliency and low runtime. The second method enables fingerprinting at all level of design process, is applicable to an arbitrary optimization step, and produces numbers of distinct solutions with high quality. The key idea is to superimpose additional constraints on the problem formulation so to guarantee that the final solution can be in a straightforward way translated into k different high quality solutions. We have implemented this on the NP-complete GC problem and tested on a number of standard benchmarks. Fingerprinting random graphs introduces overhead, while for graphs generated from real-life register allocation problems, we have successfully created millions of distinct optimal solutions with no run-time overhead.