An efficient grouping method and error probability analysis for RO-PUFs. Issue 49 (March 2015)

• Record Type: Journal Article
• Authors: Kömürcü, Giray; Pusane, Ali Emre; Dündar, Günhan
• Is Part Of: Computers & security. Issue 49(2015)
• Page Start: 123
• Page End: 131
• Publication Date: 2015-03
• Subjects: PUF -- Physical unclonable functions -- Reliability -- Robustness -- Ring oscillator -- FPGA; Computer security -- Periodicals; Electronic data processing departments -- Security measures -- Periodicals; 005.805
• DOI: 10.1016/j.cose.2014.11.009
• Languages: English
• ISSNs: 0167-4048

Abstract

Abstract: Physical Unclonable Functions (PUFs) are primitives that have wide usage areas in information security. Ordering based Ring Oscillator (RO)-PUFs have been introduced recently to overcome the robustness and area efficiency issues related to PUF implementations. With this approach, 100% robust outputs are generated, providing a solution for cryptographic key generation. High entropy extraction with relatively few ROs is also achieved, resulting in high area utilization of the PUF circuit. Frequency threshold determination is the most critical step in ordering based RO-PUFs, and determines a trade-off between area efficiency and robustness. In this work, we overview an efficient grouping method for RO-PUFs and analyze the error vulnerability of PUFs based on the frequency threshold determination. Next, we analyze the length of groups used in such PUF circuits and determine the symbol error probability. In addition to these, we demonstrate the relationship between the symbol error probability and bit error probability. We also investigate the bit error probability based on the wrong determination of the frequency threshold in ordering based RO-PUFs. Finally, a trade-off between area usage and robustness is presented for identification applications. Highlights: We analyze the error vulnerability of PUFs based on the frequency threshold. We analyze the length of groups and determine the symbol error probability. We demonstrate the relationship between the symbol and bit error probability. We investigate the bit error probability in ordering based RO-PUFs. A trade-off between area and robustness is presented for identification applications.