| Info |
|---|
Title of the paper: Efficient helper data reduction in SRAM PUFs via lossy compression Available at: https://ieeexplore.ieee.org/abstract/document/8342240 |
...
The paper gives a good motivation for using SRAM PUF
The challenge of using SRAM PUF-based key generation on FPGAs is that high-capacity NVM, such as Flash, is not available on chip. Only expensive one-time-programmable (OTP) memory with limited capacity, such as e-fuses, can be utilized to store helper data. ā Relevant to SCuM too: no NVM
The goal of this work is to reduce the size of the helper data using lossy compression, in order to decrease the required OTP memory size.
The authors distinguish two types of helper data: helper data due to
(1) the syndrome (ECC: repetition code + BCH) and
(2) the reliability mask (pre-selection)
Lossless compression techniques have been proposed to reduce HDS for both code syndrome [12] and reliability mask [13]. The ā1-out-of-nā method [20] is able to decrease the overall BER and thus HDS by sacrificing a large number of raw bits.
...
the lossy compression method achieves a 30% reduction in mask size compared to lossless compression, at the cost of doubling the number of raw bits. Including the syndrome helper data, the algorithms achieves a 25% reduction in total HDS, requiring 206 bits.
lower min-entropy rate does not present a challenge to the application of this scheme
This work focus on the trade-off between number of raw SRAM PUF bits and helper data size for both syndrome and reliability mask.