PUF Enrollment and Life Cycle Management: Solutions and Perspectives for the Test Community

Abstract

Physically Unclonable Functions (PUFs) allow to extract unique fingerprints from silicon chips. The applications are numerous: chip identification, chip master key extraction, authentication protocol, unique seeding, etc. However, secure usage of PUF requires some precautions. This paper reviews industrial concerns associated with PUF operation, including those occurring before and after market. Namely, starting from PUF "secure" specifications, aligned with state-of-the-art standards, we explore innovative techniques to handle enrollment and subsequent PUF queries, in nominal as well as in adversarial environment.

Sub-Space Modeling: An Enrollment Solution for XOR Arbiter PUF using Machine Learning

“…Our proposed solution can be used in emerging authentication and key generation protocols which demand to generate an equivalent predictive model of a target XOR arbiter PUF on a trusted third party (TTP) verifier server [1]- [3]. This solution should collectively reduce the number of CRPs in order to obtain an accurate model of XOR Arbiter PUF with large XOR size.…”

Strong PUF Enrollment with Machine Learning: A Methodical Approach

“…Security in internet-based communication, especially in IoT systems and Cyberphysical systems, is facing new opportunities and challenges as concepts such as Physically Unclonable Function (PUF)-based cryptographic methods are emerging. PUF is considered as one of the emerging security primitives for resource-constraint ecosystems in the field of IoT [1,2]. PUF is characterized as a hardware bound function which utilizes the unit-specific micro-variations to generate device-specific data.…”

“…To deal with the shortage of storage space to contain all the CRPs of a strong PUF, it is potential to generate an estimation of strong PUF using Machine Learning modeling [2]. There already exists a matured research field for strong PUF modeling, which contains a large number of different techniques that tackle the PUF modeling in different ways [11][12][13][14][15][16].…”

An Efficient Approach to Model Strong PUF with Multi-Layer Perceptron using Transfer Learning

“…The fundamental idea of PUF is to implement a hardware infrastructure that leverages the intrinsic manufacturing variations (e.g., threshold voltage, critical dimensions) to extract device specific fingerprints. PUFs are potential hardware primitives for light-weight device authentication and encryption key generation [1]- [4].…”