Digital Fingerprint: A Practical Hardware Security Primitive

Abstract

Digital fingerprinting was introduced for the protection of VLSI design intellectual property (IP). Since each copy of the IP will receive a distinct fingerprint, it can also be used as an identification for the IP or the integrated circuits (IC). This enables the IP/IC designer to trace each piece of the IP/IC and thus identify the dishonest user should piracy or misuse occurs. In this chapter, after defining the basic requirements of fingerprinting, we focus on how to solve the core challenge in digital fingerprinting, namely, how to effectively create large amount of distinct but functionally identical IPs. We first use the graph coloring problem as an example to demonstrate a general approach based on constraint manipulation; then we show how the popular iterative improvement paradigm can be leveraged for fingerprinting; the highlight will be three recently developed post-silicon fingerprinting techniques that can be automatically integrated into the design and test phases: the first two approaches take advantages of the Observability Don't Cares and Satisfiability Don't Cares, which are almost always present in IC designs, to generate fingerprints. The third method utilizes the different interconnect styles between flip flops in a scan chain to create unique fingerprints that can be detected with ease. These techniques have high practical values.