Manipulating Manufacturing Variations for Better Silicon-Based Physically Unclonable Functions

Physically Unclonable Functions (PUFs) provide interesting solutions to tnany security related issues. For instance, silicon-based PUFs are novel circuits that exploit manufacturing variations to extract unique signatures from chips. Such signatures are convenient for chip authentication and cryptographic key generation. Since variations are typically detrimental to ICs, a great deal of research is geared towards suppressing them. However, in the case of PUFs, it has been shown that wily systematic manufacturing variations are harmful and random manufacturing variations are actually the source of PUF quality. In this paper, we investigate two techniques that manipulate manufacturing variations to improve PUFs: (i) a cell layout technique that reduces systematic variation; (ii) a design technique that increases random variation. Results show that the layout technique improves PUF uniqueness by as much as 14% and the design technique improves PUF reliability by as much as 25%.