Innovatrics says its latest submission to the National Institute of Standards and Technology’s Evaluation of Latent Fingerprint Technologies (ELFT) benchmark places its latent fingerprint algorithm among the most accurate globally for rank-1 performance, based on results published by NIST near the start of this month.
ELFT evaluates the accuracy of latent fingerprint matching using features marked by experienced human latent fingerprint examiners. The program is designed to measure the state of the art in feature-based latent matching by comparing searches performed with images alone against searches that use different feature sets marked by examiners.
Rank-1 Performance in Operational Latent Searches
In the FBI Provided Solved Dataset #1, Innovatrics reports a rank-1 hit rate of 98.6 percent on 285 operational probes collected from a specific crime category. The company frames rank-1 performance as operationally important because it can reduce time spent on manual comparisons by quickly surfacing the correct subject at the top of the candidate list, allowing examiners to move through casework more efficiently.
The rank-1 hit rate is derived from the false negative identification rate (FNIR) and is commonly used in fingerprint laboratories where candidate lists returned by automated systems are reviewed manually. In that workflow, the hit rate reflects how often the correct subject appears at or above a given rank, with rank-1 representing the most direct path to a usable candidate for review.
Innovatrics previously highlighted strong performance in the ELFT program in a December 2024 results release, and the latest figures position the company as seeking incremental gains in the rank-1 category as ELFT continues to evolve as a reference point for latent identification accuracy.
Threshold-Based Search and Cost Drivers
Beyond rank-based results, Innovatrics points to its standing in “search with threshold,” a category that assesses the balance between FNIR and false positive identification rate (FPIR) when non-matching candidates are automatically filtered based on similarity score. In that operating model, a higher threshold can reduce the volume of candidates requiring manual review, while increasing the importance of maintaining low false negatives so the correct subject is still returned.
In the DoD Dataset #1, described as the most recent and largest dataset with 5,259 operational probes collected by the United States Department of Defense, Innovatrics says its algorithm remains among the most accurate in both rank-based and threshold-based search. The company positions threshold-based performance as closely tied to operating cost because the two error rates influence how much work is pushed to human review.
The company has repeatedly used NIST programs as a validation channel across modalities, including a MINEX III update focused on fingerprint interoperability and accuracy and a broader positioning of its portfolio in NIST testing as part of its multi-modality strategy.
While the ELFT results are often used as a comparative reference for algorithm quality, NIST notes that published results do not constitute an endorsement of any specific product or vendor by the U.S. government.
Innovatrics has been expanding its commercial footprint across both public-sector and regulated private-sector identity workflows, including recent work on onboarding controls such as active liveness for regulated digital onboarding and integrations aimed at operationalizing biometric checks in fintech environments such as its latest partnership with Payout.
–
By Cass Kennedy and Alex Perala
