The PUFs we attacked successfully include standard Arbiter PUFs and Ring Oscillator PUFs of arbitrary sizes, and XOR Arbiter PUFs, Lightweight Secure PUFs, and Feed-Forward Arbiter PUFs of up to a given size and complexity. The attacks are based upon various machine learning techniques, including a specially tailored variant of Logistic Regression and Evolution Strategies.
Our results were obtained on a large number of CRPs coming from numerical simulations, as well as four million CRPs collected from FPGAs and ASICs. The performance on silicon CRPs is very close to simulated CRPs, confirming a conjecture from earlier versions of this work. Our findings lead to new design requirements for secure electrical PUFs, and will be useful to PUF designers and attackers alike.Category / Keywords: implementation / implementation / Physical Unclonable Functions, Machine Learning, Cryptanalysis, Physical Cryptography Date: received 25 Feb 2013, last revised 20 Aug 2013 Contact author: ruehrmair at in tum de Available format(s): PDF | BibTeX Citation Version: 20130820:115911 (All versions of this report) Discussion forum: Show discussion | Start new discussion