# Preprint presents BioBind protocol for biometric zero-knowledge authentication

Published 2026-11-10 · Crypto.world Team · Crypto.world

> BioBind links biometric deduplication at enrollment with zero-knowledge authentication when credentials are used in decentralized applications. Its prototype results include tests of enrollment and authentication, while the paper says full backend integration remains open.

## A protocol spanning enrollment and use

A new preprint describes BioBind, a protocol that connects biometric deduplication during enrollment with zero-knowledge authentication when credentials are used in decentralized applications. The paper’s authors are Taotao Wang, Xiaoxiao Wu and Shengli Zhang. The design aims to resist repeated enrollment while supporting private, publicly verifiable authentication.

Source [arxiv.org](https://arxiv.org/html/2611.12636)

The paper frames the problem around decentralized applications that use identity credentials to limit repeated participation or provide one-per-person benefits without exposing a user’s identity at each interaction. It distinguishes two biometric tasks: one-to-many deduplication checks a new enrollee against an existing gallery, while one-to-one authentication checks a later probe against an enrolled template. BioBind asks how a deduplication result can become an enrollment record that later private authentication can rely on.

In the model described in the paper, biometric capture and large-scale private deduplication take place off chain, while smart contracts authorize credential issuance and use. Those stages span separate trust domains, and a correct deduplication verdict covers only the inputs checked and the gallery snapshot used for the check. The enrollee may control its own host, submitted data and request timing.

Without additional binding and state checks, concurrent enrollments could issue duplicate credentials even when the verdicts are correct, while substituted inputs could authorize unrelated credentials. One example in the paper is an enrollee querying the gallery with its captured template, receiving a valid unique verdict, then inserting an encrypted zero vector instead. The person would have no entry in the gallery and could pass a later deduplication check again.

## Checks across the backend

BioBind’s four backend interface conditions tie the deduplication query, gallery insertion and issued commitment to one captured vector; they also authenticate gallery transitions and require a sound uniqueness verdict at finalization. The paper presents these conditions as safeguards across separate stages, rather than treating a correct result from one check as sufficient for the whole process.

A second example is a snapshot race. Two enrollment sessions opened by the same person can each check a gallery snapshot that lacks that person and correctly return a unique result. If both sessions then insert entries and issue credentials, the person enrolls twice. The paper also describes a dropped-merge failure: a compute node can discard a new encrypted gallery entry yet report an updated commitment, leaving the person absent from the gallery.

For its encrypted instance, the paper specifies bound requests, rejection of stale sessions and committee-confirmed migration. Its capture-bound authentication links credential use to the enrolled template and to a fresh probe originating from a device and bound to the challenge. The paper gives a multi-session composition argument, conditional on explicit assumptions about capture, the backend and cryptography.

The paper situates these checks alongside earlier systems that address parts of the process. It describes World ID’s published multiparty-computation protocol and BioZKFHE as approaches to private one-to-many deduplication, and notes that other work proves a match against a committed template and verifies it on chain. Those approaches do not, by themselves, establish the full link between the capture, the gallery entry, the credential and a uniqueness verdict that remains valid when enrollment is finalized.

## Prototype results and limits

The prototype uses LFW face features and rejects the composition attacks tested in the paper. Its unlinkable authentication circuit uses 31,445 constraints and about 305,000 gas. For an equal-public-field baseline, median witness-plus-proving time is 1.989 seconds for verifiable matching and 1.119 seconds when matching is trusted to the device.

In separate encrypted-enrollment tests, three histories of 16,384 registrations each completed without decoding errors at the tested checkpoints. The paper also reports a limit tied to biometric accuracy: on the Bosphorus database, cooperative subjects who changed appearance evaded deduplication in 1.8% of re-enrollment attempts at the operating threshold. The authors write, “Full backend integration remains open.”

## Preprint details

The preprint was submitted on Nov. 9, 2026, and lists 29 pages, seven figures and 19 tables.

## Sources

1. Original source · Preprint [arXiv](https://arxiv.org/html/2611.12636) arXiv:2611.12636

## Related stories

- [GRIDSHIELD Preprint Proposes Order-Independent Clearing for P2P Energy Trading](https://crypto.world/news/gridshield-manipulation-resistant-p2p-energy-trading)
- [Preprint renames LinkDID as Braid, an identity design that links identifiers across apps](https://crypto.world/news/braid-revises-cross-app-identity-design)
- [Preprint proposes one-transaction proofs for Ethereum quantum migration](https://crypto.world/news/researchers-propose-one-transaction-proofs-ethereum-quantum-migration)
- [Paper Finds ZK Rewards Raised Attackers’ Share in Federated Learning Tests](https://crypto.world/news/paper-finds-zk-rewards-raised-attackers-share-federated-learning)
- [Researchers test Besu system for private local energy settlement](https://crypto.world/news/researchers-test-besu-private-local-energy-settlement)

## Explore

[More articles](https://crypto.world/news)

---

Canonical: https://crypto.world/news/preprint-biobind-biometric-zero-knowledge-authentication
