Skip to main content

Concepts

What is a Biometric Template?​

A biometric template is a compact mathematical representation of a biometric measurement. An iris code, for example, might encode the texture pattern of an iris as a bit vector of 11k bits. The actual bit length depends on the feature extraction, whether it is iris, face or palm scans. All share that they are fixed-length vectors designed to make comparison cheap and reliable.

The challenge is that these vectors are sensitive and irrevocable. Unlike a password hash, a leaked biometric template cannot be rotated. Any system that stores or transmits raw templates creates a permanent liability.

Secret Sharing​

Secret sharing splits a value into nn shares such that any subset of size <t< t reveals nothing about the original value, while any subset of size ≥t\geq t can reconstruct it. In TACEO:Match, each enrolled template is split into shares distributed across independent MPC nodes on the TACEO Network. No node holds a full template.

For a deeper treatment of MPC and the secret-sharing primitives TACEO uses, see the MPC Resources page.

The Matching Function​

Biometric matching reduces to computing a distance metric between two templates and checking whether it falls below a threshold:

match = distance(probe, enrolled) < τ

For iris codes, this is Hamming distance - the fraction of bit positions that differ. Is the Hamming distance below a certain threshold ττ we can deduce that two scans belong to the same person.

TACEO:Match evaluates this distance function directly on the secret shares of the enrolled template, without reconstructing the template at any node. The same approach extends to other distance-based biometric representations (e.g. cosine similarity for face embeddings) by implementing the corresponding comparison circuit.

Composition with Nullifiers​

TACEO:Match and Distributed Nullifiers are complementary primitives built on the same threshold MPC stack. Nullifiers derive a deterministic, unlinkable identifier from a secret input via OPRF; Match verifies that a biometric probe corresponds to an enrolled identity via MPC matching.

A common composition: use Match to verify biometric uniqueness, then issue a nullifier tied to the verified identity. The user gets both sybil resistance (from Match) and unlinkability across interactions (from the nullifier).