| Title |
A XOR-AND Transpose Physical Unclonable Function That Allows Selection of Bitwise Operators and Input Positions |
| Authors |
백창재(Changjae Beak) ; 박서현(Seo Hyeon Park) ; 이정원(Jungwon Lee) ; 선우경(Wookyung Sun) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.12.38 |
| Keywords |
Physical unclonable function; PUF; FPGA; Machine learning attack |
| Abstract |
The importance of establishing security for device-to-device or device-to-server communication without human intervention emerges as the IoT technology has grown to be an integral part of everyday life. However, vulnerabilities were found in software security technology, and Physical Unclonable Function (PUF) was proposed as a solution with stronger security. PUF uses deviations from the process to generate challenge-response pairs (CRP). PUF itself is strong, but it is vulnerable to machine learning attacks if an attacker knows some of the CRP values. This paper proposes a XOR-AND Transpose-PUF(XAT-PUF) that increases the number of CRP cases by using a ROW-COL selector for I/O structures and making it possible to select bit operators. In addition, Multi Combination-XAT(MC-XAT) expressed in a multi-bit structure is also proposed. The simulation results that the proposed structure improved by an average of 258% compared to MA-SBC and 83% compared to MC-SBC in the improvement rate of machine learning attack defense performance. In addition, the proposed structure is implemented up to the resource consumption performance using FPGA. |