On fuzzy syndrome hashing with LDPC coding

The last decades have seen a growing interest in hash functions that allow some sort of tolerance, e.g. for the purpose of biometric authentication. Among these, the syndrome fuzzy hashing construction allows to securely store biometric data and to perform user authentication without the need of sharing any secret key. This paper analyzes this model, showing that it offers a suitable protection against information leakage and several advantages with respect to similar solutions, such as the fuzzy commitment scheme. Furthermore, the design and characterization of LDPC codes to be used for this purpose is addressed.

💡 Research Summary

The paper addresses the problem of securely storing and authenticating biometric data without a secret key, a task complicated by the inherent variability of biometric measurements. Traditional fuzzy commitment schemes store a hash of a random codeword together with a “shift” vector (the difference between the biometric template and the codeword). While this approach enables tolerant authentication, the stored shift vector can leak information about the original biometric template, especially when parts of the biometric are highly predictable.

To mitigate this leakage, the authors propose a “fuzzy syndrome hashing” construction. Instead of storing a random codeword, the system stores the syndrome of the biometric vector with respect to a linear error‑correcting code. The stored data consist of (Hₐ(x), H x), where Hₐ is a conventional cryptographic hash and H x is the syndrome obtained from the parity‑check matrix H of an