Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A novel reliability attack of Physical Unclonable Functions

Gaoxiang Li, Yu Zhuang

TL;DR

This work examines how reliability-based ML attacks threaten PUF-based security and investigates defenses via majority voting. It reveals that majority voting can significantly reduce attack efficacy against existing methods, then introduces a low-dimension high-fidelity (LDHF) reliability representation that preserves unreliability information while reducing output dimensionality, enabling neural networks to crack highly reliable PUFs. Through FPGA experiments and data-driven analysis, the authors show LDHF-based attacks can overcome MV defenses in many scenarios, highlighting a new class of vulnerabilities in high-reliability PUF designs. The results motivate rethinking PUF robustness and point to advanced representations and architectures as essential tools for assessing and improving hardware security in IoT devices.

Abstract

Physical Unclonable Functions (PUFs) are emerging as promising security primitives for IoT devices, providing device fingerprints based on physical characteristics. Despite their strengths, PUFs are vulnerable to machine learning (ML) attacks, including conventional and reliability-based attacks. Conventional ML attacks have been effective in revealing vulnerabilities of many PUFs, and reliability-based ML attacks are more powerful tools that have detected vulnerabilities of some PUFs that are resistant to conventional ML attacks. Since reliability-based ML attacks leverage information of PUFs' unreliability, we were tempted to examine the feasibility of building defense using reliability enhancing techniques, and have discovered that majority voting with reasonably high repeats provides effective defense against existing reliability-based ML attack methods. It is known that majority voting reduces but does not eliminate unreliability, we are motivated to investigate if new attack methods exist that can capture the low unreliability of highly but not-perfectly reliable PUFs, which led to the development of a new reliability representation and the new representation-enabled attack method that has experimentally cracked PUFs enhanced with majority voting of high repetitions.

A novel reliability attack of Physical Unclonable Functions

TL;DR

This work examines how reliability-based ML attacks threaten PUF-based security and investigates defenses via majority voting. It reveals that majority voting can significantly reduce attack efficacy against existing methods, then introduces a low-dimension high-fidelity (LDHF) reliability representation that preserves unreliability information while reducing output dimensionality, enabling neural networks to crack highly reliable PUFs. Through FPGA experiments and data-driven analysis, the authors show LDHF-based attacks can overcome MV defenses in many scenarios, highlighting a new class of vulnerabilities in high-reliability PUF designs. The results motivate rethinking PUF robustness and point to advanced representations and architectures as essential tools for assessing and improving hardware security in IoT devices.

Abstract

Physical Unclonable Functions (PUFs) are emerging as promising security primitives for IoT devices, providing device fingerprints based on physical characteristics. Despite their strengths, PUFs are vulnerable to machine learning (ML) attacks, including conventional and reliability-based attacks. Conventional ML attacks have been effective in revealing vulnerabilities of many PUFs, and reliability-based ML attacks are more powerful tools that have detected vulnerabilities of some PUFs that are resistant to conventional ML attacks. Since reliability-based ML attacks leverage information of PUFs' unreliability, we were tempted to examine the feasibility of building defense using reliability enhancing techniques, and have discovered that majority voting with reasonably high repeats provides effective defense against existing reliability-based ML attack methods. It is known that majority voting reduces but does not eliminate unreliability, we are motivated to investigate if new attack methods exist that can capture the low unreliability of highly but not-perfectly reliable PUFs, which led to the development of a new reliability representation and the new representation-enabled attack method that has experimentally cracked PUFs enhanced with majority voting of high repetitions.
Paper Structure (27 sections, 3 equations, 6 figures, 10 tables, 1 algorithm)

This paper contains 27 sections, 3 equations, 6 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background on Arbiter PUF and its variants
  3. Arbiter PUF
  4. XOR-Arbiter-PUF
  5. Interpose-PUF
  6. Brief Survey of Reliability-based Machine Learning Attacks
  7. PUF Circuit Architecture-Tailored Attack Methods
  8. Generally Applicable Reliability-based Attack Methods
  9. A Multiclass-based Attack Method
  10. The Multi-label Multi-side-channel-based Method
  11. The Auxiliary Learning Side-channel Attack Method
  12. Experimental Comparative Study of Existing Methods
  13. Experimental Tools and Setup
  14. The Experimental Attack Study
  15. A Simplistic Defense Against Existing Reliability-Based Attacks
  16. ...and 12 more sections

Figures (6)

  • Figure 1: An arbiter PUF with n bits of challenge.
  • Figure 2: Illustration of an XOR-PUF with n arbiter PUF components; the final response is the XORed result of n arbiter-PUF responses.
  • Figure 3: Schematic representation of Interpose PUFs.
  • Figure 4: Illustration of the neural network architectures of the MLMSA gao2023mlmsa (left) and the ALScA 9973338 (right)
  • Figure 5: PUF reliability performance under different numbers of majority votes applied
  • ...and 1 more figures