This paper deals with design of Physical Unclonable Functions (PUFs) based on FPGA. The goal was to propose a cheap, efficient and secure device identification or even a cryptographic key generation based on PUFs. Therefore, a proposal of a ring oscillator (RO) based PUF producing more output bits from one RO pair is presented. 24 Digilent Basys 2 FPGA boards were tested and statistically evaluated indicating suitability of the proposed design for device identification.

No abstract

No abstract

The Montgomery inverse is used in cryptography for the computation of modular inverse of b modulo a, where a is a prime. We analyse existing algorithms from the point of view of their hardware implementation. We propose a new, hardware-optimal algorithm for the calculation of the classical modular inverse. The left-shift binary algorithm is shown to naturally calculate the classical modular inverse in fewer operations than the algorithm derived from the Montgomery inverse.

Secret keys are usually stored in a nonvolatile memory, which can be hard to secure. An alternative is to generate the keys "on-the-fly" by using the inherent uniqueness of a device based on the manufacturing process variations. This is realized by physical unclonable functions (PUFs). A promising approach is to construct an intrinsic PUF based on SRAM memory, since many electronic devices have embedded SRAM. However, using a SRAM as PUF requires the stability of the SRAM fingerprint under a wide variety of conditions, and the SRAM fingerprint must be unique.In this paper, we show that a 16Kbyte SRAM fingerprint contains sufficient entropy to uniquely identify each chip. In addition, if a postprocessing error correction is applied, the fingerprint can be used to generate a stable 4Kbit key.