Modeling Attack Security of Physical Unclonable Functions based on Arbiter PUFs by Nils Wisiol

Author:Nils Wisiol
Language: eng
Format: epub
ISBN: 9783031292071
Publisher: Springer International Publishing

(4.1.1)

4.1.1 Pseudorandom Input Transformation

We demonstrate the influence of input transformations on the learning hardness of logistic regression attacks in Fig. 4.1. To contrast the classic design, where all arbiter chains receive the same challenge, we implemented a simulation of XOR Arbiter PUFs with pseudorandom sub-challenge generators, where all arbiter chains receive an individual pseudorandom challenge chosen by seeding the generator with the master-challenge and the index of the sub-challenge. For our implementation, we used a pseudorandom generator based on the Mersenne Twister. Assuming security of the pseudorandom generator, we can guarantee that the sub-challenges are chosen indistinguishable from truly random sub-challenges and feature vectors (for all polynomially time-bounded observers, i.e., including the machine learning attacker).

Fig. 4.1Success rate of logistic regression attacks on simulated XOR Arbiter PUFs with 64-bit arbiter chains and four arbiter chains each, based on at least 250 samples per data point shown. Accuracy better than 70% is considered success, but we only observe accuracy around 50 and 99%. Four different designs are shown: of the four arbiter chains in each instance, an input transform is used that transforms zero, one, two, and three challenges pseudorandomly, keeping the remaining challenges unmodified

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.