Identities, Not Pairs: Distribution-Free False-Accept Control for Embedding-Based Verification
Abstract
Embedding-based verification often calibrates a false-accept threshold from all impostor pairs among m identities, although these pairs share identities. We characterize a worst-case limit: when hub identities score highly against a fraction ρ of the population, a calibration-only threshold with marginal false-accept rate (FAR) at most α under every identity-exchangeable law reaches the hub score on hub-free calibration sets with probability at most (m+2)α/(2ρ). Uniformly accepting near that score at levels well below ρ/m therefore requires trial information. We propose COVE, which ranks the trial pair among all N=binomm+22 pairs formed by augmenting the calibration units with the probe and reference. Under identity exchangeability, its marginal FAR is at most ⌊αN⌋/N, with first possible acceptance at 1/N=O(1/m2). The rule also retains this guarantee with a fixed symmetric cohort-normalized score using an independent cohort. On 6 benchmarks and 8 frozen encoders, realised FAR was consistent with the exact bound. Where both valid calibration-only baselines accepted nothing, COVE retained a median 100% of the sampled oracle's true-accept rate over reliable comparisons. In contrast, an empirical quantile over dependent pairs reached 2.87α FAR, and a binomial upper confidence bound over pairs missed its nominal 95\% conditional-FAR target in up to 55\% of calibration draws. Symmetric CSLS raised COVE's true-accept rate by up to 4.3-fold in the tested settings. The code is available in https://github.com/salimkhazem/Cove.git
Related
- Safety Verification for Voltage Control in Inverter Dominated Radial Distribution Networks under Actuator and Sensor Adversarial Attacks
- quantum-group: Exact symbolic verification of R-matrix and Yang--Baxter identities for Uq(sl2) and GLq(2|1)
- The kernel mean embedding of a random probability distribution and its principal components
- Embedding Information in Policy Randomness: Distribution-Preserving Implicit Communication in MDPs
- Safe Control of Semi-Explicit Differential-Algebraic Systems: Control Barrier Functions with SOS Verification
BibTeX
@misc{khazem2026identities,
title = {{Identities, Not Pairs: Distribution-Free False-Accept Control for Embedding-Based Verification}},
author = {Salim Khazem},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.35277},
eprint = {2610.35277},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.35277},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Identities, Not Pairs: Distribution-Free False-Accept Control for Embedding-Based Verification
AU - Salim Khazem
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2610.35277
UR - https://arxiv.org/abs/2610.35277
AB - Embedding-based verification often calibrates a false-accept threshold from all impostor pairs among m identities, although these pairs share identities. We characterize a worst-case limit: when hub identities score highly against a fraction ρ of the population, a calibration-only threshold with marginal false-accept rate (FAR) at most α under every identity-exchangeable law reaches the hub score on hub-free calibration sets with probability at most (m+2)α/(2ρ). Uniformly accepting near that score at levels well below ρ/m therefore requires trial information. We propose COVE, which ranks the trial pair among all N=binomm+22 pairs formed by augmenting the calibration units with the probe and reference. Under identity exchangeability, its marginal FAR is at most ⌊αN⌋/N, with first possible acceptance at 1/N=O(1/m^2). The rule also retains this guarantee with a fixed symmetric cohort-normalized score using an independent cohort. On 6 benchmarks and 8 frozen encoders, realised FAR was consistent with the exact bound. Where both valid calibration-only baselines accepted nothing, COVE retained a median 100% of the sampled oracle's true-accept rate over reliable comparisons. In contrast, an empirical quantile over dependent pairs reached 2.87α FAR, and a binomial upper confidence bound over pairs missed its nominal 95\% conditional-FAR target in up to 55\% of calibration draws. Symmetric CSLS raised COVE's true-accept rate by up to 4.3-fold in the tested settings. The code is available in https://github.com/salimkhazem/Cove.git
ER -