Skip to main content
eScholarship
Open Access Publications from the University of California

Digital Physical Unclonable Functions: Architecture and Applications

  • Author(s): Xu, Teng
  • Advisor(s): Potkonjak, Miodrag
  • et al.
Abstract

The rapid growth of small form, mobile, and remote sensor network systems require secure and ultralow power data collection and communication solutions due to their energy constraints. The physical unclonable functions (PUFs) have emerged as a popular modern security primitive. They have the property of low power/energy, small area, and high speed. Moreover, they have excellent security properties and are resilient against physical and side-channel attacks. However, traditional PUFs have two major problems. The first is that current designs are analog in nature and lack susceptibility in environmental and operational conditions, e.g., supply voltage and temperature. The second is that due to the analog nature, the analog PUFs are difficult to be integrated into existing digital circuitry.

Therefore, in this thesis, we propose the digital PUF, as a new type of security primitive. It preserves all the good properties of traditional analogy PUFs and is stable in the same sense that digital logic is stable. It has a small footprint, a small timing overhead, a low energy consumption, and can be easily integrated into existing designs. The key observation is that for any analog delay PUF, there is a subset of challenge inputs for which the PUF output is stable regardless of operation and environmental conditions. We use only such stable inputs to initialize the look-up tables (LUTs) in digital bimodal functions (DBFs) that are configured in such a way that the digital PUF is formed. We first demonstrate the concept and the FPGA-based architecture of the digital PUF. Then we present our security analysis on digital PUFs using standard randomness tests and confusion and diffusion analysis. Finally, we address security protocols of digital PUF: public key communication, and remote trust.

Main Content
Current View