A PUF-Based Key Storage Scheme Using Fuzzy Vault

Jinrong Yang, Shuai Chen, Yuan Cao2023 · DOI

TL;DR

This paper proposes a novel method for securely storing secret keys in IoT devices by leveraging Physical Unclonable Functions (PUFs) and the Fuzzy Vault cryptographic technique. Instead of traditional encryption, it directly uses the unique, noisy responses from PUFs to lock and unlock secret data, offering a hardware-based, low-cost security solution.

What Problem Does This Paper Address?

The Internet of Things (IoT) is rapidly expanding, bringing immense benefits but also significant security and privacy concerns. A critical challenge is securely storing secret keys within these devices, especially resource-constrained ones. Traditional methods often rely on Non-Volatile Memory (NVM), which can be vulnerable to physical attacks and expensive to secure. Furthermore, the process of generating and embedding keys during manufacturing can lead to supply chain security issues. Physical Unclonable Functions (PUFs) offer a promising alternative, using unique physical characteristics of integrated circuits to generate keys. However, PUFs are inherently unreliable due to aging and environmental factors. Existing solutions to this unreliability, like helper data algorithms, add complexity and overhead. This paper addresses the need for a simpler, more integrated, and hardware-rooted security solution for key storage in IoT devices, directly utilizing the fuzzy nature of PUF responses.

Methods

The core of this paper's contribution lies in its innovative application of the Fuzzy Vault scheme to PUF responses. Unlike traditional Fuzzy Vault applications that use biometrics, this work uses PUF responses as the "fuzzy biometric template."

Here's a breakdown of their approach:

  • Direct PUF Response Utilization: Instead of using PUF responses to generate a reproducible key that then encrypts data (the typical helper data approach), this method directly uses the PUF responses to lock the secret data itself.
  • Fuzzy Vault Integration: The Fuzzy Vault scheme is employed to handle the inherent noise and unreliability of PUF responses.
  • Locking Phase: A secret key is first used to construct a polynomial. This polynomial is then evaluated using PUF responses to generate a set of "real points." These real points, along with randomly generated "chaff points," form the vault, which is stored in NVM.
  • Unlocking Phase: When the secret key needs to be retrieved, a fresh PUF response is queried. This response is used to identify the real points within the vault. Due to the error-tolerant nature of the Fuzzy Vault (specifically, using Lagrange interpolation), even with noisy PUF responses, enough stable points can be reconstructed to recover the original polynomial and thus the secret key.
  • Weak vs. Strong PUF Handling: The paper proposes distinct strategies for weak and strong PUFs:
  • Weak PUFs: Each PUF instance can protect a limited amount of data. The approach involves extending the secret data and using error-correcting codes (like CRC-16) for integrity checks.
  • Strong PUFs: These have a vast number of challenge-response pairs (CRPs). This allows a single strong PUF instance to protect a much larger amount of secret data by generating multiple vaults. The paper also introduces preprocessing steps for strong PUF responses to mitigate biases and improve security.
  • Error Tolerance: The Fuzzy Vault's inherent error tolerance is crucial. It allows the system to function even if some PUF responses are unreliable, as long as a sufficient number of stable points can be reconstructed.

Datasets

Not discussed in the paper. The authors evaluate their scheme using existing weak and strong PUF implementations.

Key Insights and Ideas

  • PUFs as Direct Storage: The most significant insight is the shift from using PUFs solely for key generation to directly using their fuzzy responses for data storage. This bypasses the need for traditional encryption algorithms and helper data schemes, simplifying the overall architecture.
  • Fuzzy Vault's Natural Fit: The inherent fuzziness and unreliability of PUF responses make them a perfect match for the error-tolerant nature of the Fuzzy Vault. This is a novel application of the Fuzzy Vault technique.
  • Weak vs. Strong PUF Tradeoffs: The paper clearly delineates how the characteristics of weak and strong PUFs influence their suitability for this storage scheme, highlighting the scalability advantages of strong PUFs for protecting larger amounts of data.
  • Security Through Obscurity and Error Tolerance: The security relies on a combination of the difficulty for an adversary to distinguish real points from chaff points and the error-tolerant reconstruction of the secret data, even with noisy PUF responses.

Key Results

The paper demonstrates the feasibility and effectiveness of their proposed PUF-based Fuzzy Vault scheme through experimental evaluations:

  • High Unlocking Rate: With parameter adjustments, the proposed scheme can achieve a 100% unlocking rate.
  • Fast Locking and Unlocking: Locking times are reported to be less than 1 second, and unlocking times are in the few seconds range.
  • Weak PUF Reliability: Experiments with a weak PUF (16x16 current comparator) showed that even under varying temperatures and voltages, a significant number of reliable groups (points) could be generated, enabling successful unlocking.
  • Strong PUF Performance: Evaluations with a strong PUF (64-bit APUF) also demonstrated good reliability and the potential to protect a large number of secret data points.
  • Tradeoff Analysis: The paper provides a detailed discussion on the tradeoffs between security, reliability, and runtime, showing how parameters like polynomial degree, number of real points, and chaff points can be tuned for different application requirements.

Limitations and Future Work

  • Security Against Advanced Attacks: While the paper discusses brute-force and correlation attacks, it acknowledges that strong PUFs can be vulnerable to modeling attacks if their CRPs are exposed. The authors suggest using erasable PUF techniques and secure hardware boundaries (like TPMs) to mitigate this.
  • PUF Reliability: The overall reliability of the scheme is still dependent on the underlying PUF's stability. While the Fuzzy Vault offers error tolerance, extreme unreliability in the PUF could still lead to unlocking failures.
  • Parameter Tuning: Achieving optimal performance requires careful tuning of parameters (polynomial degree, number of real points, chaff points) based on the specific PUF characteristics and operating environment.
  • Future Work: The authors suggest further research into improving the efficiency of their methods, potentially through hardware accelerators. They also highlight the need for more comprehensive security analysis against a wider range of attacks.

Summary updated

BibTeX
@misc{yang2023puf,
  author = {Jinrong Yang and Shuai Chen and Yuan Cao},
  title = {{A PUF-Based Key Storage Scheme Using Fuzzy Vault}},
  year = {2023},
  doi = {10.3390/s23073476},
}
Figure