Seesaw: An Area-Optimized FPGA Viterbi Decoder for PUFs
Matthias Hiller, Leandro Rodrigues Lima, Georg Sigl · 2014
Physical Unclonable Functions PUFs are popular security primitives to provide cryptographic keys on FPGAs. However, PUFs require error correction to create reliable cryptographic keys. This work presents a highly optimized Viterbi decoder, adapted to the constraints of PUFs on FPGAs, primarily area but also low power. Our Seesaw architecture contains two block RAMs that are connected through a custom low-area data path. As main result, alternating data access patterns reduce the complexity of the data handling in the Viterbi decoder. Instead of translating through the entire trellis, we introduce a method that only operates on the last state. The new access pattern permits to store the intermediate results in block RAM and leads to a compact overall footprint with low register count. Synthesis results for one legacy and one state-of-the art FPGA, and a comparison to state-of-the-art implementations demonstrate the efficiency of our new Seesaw architecture. Our decoder requires only 65 FPGA slices and 2 block RAMs to carry out the entire Viterbi decoding for a popular (2, 1, [7]) convolutional code.