A Symmetric Keyring Encryption Scheme for Biometric Cryptosystems

Yen-Lung Lai, Jung Yeon Hwang, Zhe Jin, Soohyong Kim, Sangrae Cho, Andrew Beng Jin TeoharXiv 2018 · OpenReview

TL;DR

This paper introduces Symmetric Keyring Encryption (SKE), a novel biometric cryptosystem that treats biometric secret-binding as a fuzzy symmetric encryption problem. By leveraging a "resilient vector pair" derived from locality-sensitive hashing and Shamir's secret-sharing, SKE aims to securely bind secrets to biometrics with high accuracy for legitimate users and strong resistance against imposters.

What Problem Does This Paper Address?

Traditional cryptographic systems rely on exact, deterministic keys. However, biometrics, like fingerprints, are inherently noisy and prone to variations, making deterministic approaches unreliable. Biometric cryptosystems aim to bridge this gap by binding secrets to biometrics, but existing methods like fuzzy commitment and fuzzy vault have limitations in terms of security, resilience, or complexity. This paper addresses the challenge of creating a robust and secure biometric cryptosystem that can handle the inherent fuzziness of biometric data while offering strong privacy and security guarantees.

Methods

The core of SKE is the concept of a "resilient vector pair" (RV pair), inspired by Rivest's keyring model. This RV pair acts analogously to a public/private key pair in symmetric cryptosystems. The paper utilizes Index-of-Maximum (IoM) hashing, a form of locality-sensitive hashing, to generate these RVs from biometric vectors. IoM hashing is designed to preserve similarity between input vectors in its hashed output.

The SKE scheme involves two main phases:

  1. Enrollment: A secret is encoded using Shamir's secret-sharing scheme. Then, two RVs are generated from the biometric input using IoM hashing with different nonces. One RV is used to encrypt the secret shares, creating a "secure sketch," and an authentication tag is generated. The public helper data consists of the nonces, the secure sketch, and the tags.
  2. Secret Retrieval: A query biometric vector is used to generate a corresponding RV pair. The authentication tags are checked to filter out incorrect matches. If enough valid matches are found, Shamir's secret-sharing is used to reconstruct the secret. A key filtering mechanism is employed to ensure only genuine pairs contribute to the unlocking set.

Datasets

The experiments were conducted using subsets of the following standard fingerprint benchmark datasets:

  • FVC 2002 (subsets DB 1 and DB 2)
  • FVC 2004 (subsets DB 1 and DB 2)
  • FVC 2006 (subset DB 1)

The paper used Set B of these datasets for generating KPCA projection matrices and Set A for secret retrieval experiments.

Key Insights and Ideas

  • Biometrics as Fuzzy Symmetric Keys: Reframing the biometric secret-binding problem as a fuzzy symmetric encryption problem allows for leveraging existing cryptographic concepts in a novel way.
  • IoM Hashing for Similarity Preservation: The use of IoM hashing is crucial as it generates hashed vectors that maintain the similarity of the original biometric inputs, which is essential for successful retrieval.
  • TAG for Integrity and Filtering: The authentication TAG plays a vital role in ensuring data integrity and acting as a filter, preventing decryption with incorrect or corrupted data. This mechanism is key to achieving high resilience.
  • Strong Resilience with High Probability: The scheme demonstrates the ability to retrieve the exact secret with overwhelming probability for genuine inputs, even with noisy biometrics.
  • Tag Indistinguishability for Privacy: The proposed scheme incorporates a "tag indistinguishability" property, offering strong privacy protection by making it difficult to link different encryptions performed by the same user.

Key Results

  • High Accuracy: The empirical evaluations on FVC datasets showed that SKE can achieve high accuracy in secret retrieval, with False Rejection Rates (FRR) and False Acceptance Rates (FAR) comparable to or better than existing methods, especially for certain datasets. For instance, on FVC2002 DB1 and DB2, SKE achieved very low FAR (often 0.00%) with reasonable FRR.
  • Performance Preservation: SKE demonstrated the ability to preserve or even improve the performance (measured by Equal Error Rate - EER) compared to the original biometric system, which is a significant achievement as many biometric cryptosystems can degrade performance.
  • Security Analysis: The paper formalizes and analyzes the threat model, showing SKE's resistance to brute-force attacks, false-accept attacks, and tag indistinguishability attacks. The security is shown to be dependent on parameters like F (field size) and m (number of IoM hashed entries).
  • Efficiency: The average encoding time increases with larger k (secret size), but the decoding time for imposter attempts remains relatively constant, suggesting efficient retrieval for genuine users.

Limitations and Future Work

  • Parameter Tuning: The performance and security of SKE are sensitive to parameter choices (m, F, δ, k). Finding optimal parameters for different biometric modalities and security requirements might require further investigation.
  • Computational Complexity: While SKE offers strong security, the computational complexity of IoM hashing and polynomial interpolation can be a factor, especially for very large datasets or high security requirements.
  • Trade-off: The paper acknowledges a trade-off between system performance (accuracy) and security. Increasing security parameters might lead to a slight degradation in accuracy.
  • Future Work: The authors suggest that future research could explore applying SKE to other vectorial biometrics beyond fingerprints and further optimizing the IoM hashing and filtering mechanisms for even greater efficiency and security. They also highlight the importance of considering the non-uniformity of biometric data in future cryptographic research.

Summary updated

Abstract

In this paper, we propose a novel biometric cryptosystem for vectorial biometrics named symmetric keyring encryption (SKE) inspired by Rivest's keyring model (2016). Unlike conventional biometric secret-binding primitives, such as fuzzy commitment and fuzzy vault, the proposed scheme reframes the biometric secret-binding problem as a fuzzy symmetric encryption problem with a notion called resilient vector pair. In this study, the pair resembles the encryption-decryption key pair in symmetric key cryptosystems. This notion is realized using the index of maximum hashed vectors - a special instance of the ranking-based locality-sensitive hashing function. With a simple filtering mechanism and [m,k] Shamir's secret-sharing scheme, we show that SKE, both in theoretical and empirical evaluation, can retrieve the exact secret with overwhelming probability for a genuine input yet negligible probability for an imposter input. Though SKE can be applied to any vectorial biometrics, we adopt the fingerprint vector as a case of study in this work. The experiments have been performed under several subsets of FVC 2002, 2004, and 2006 datasets. We formalize and analyze the threat model of SKE that encloses several major security attacks.
BibTeX
@misc{lai2018symmetric,
  author = {Yen-Lung Lai and Jung Yeon Hwang and Zhe Jin and Soohyong Kim and Sangrae Cho and Andrew Beng Jin Teoh},
  title = {{A Symmetric Keyring Encryption Scheme for Biometric Cryptosystems}},
  year = {2018},
  eprint = {1809.11045},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1809.11045},
}
Figure