Innovatrics Ranks First Across Four Key NIST ELFT Accuracy Metrics

The latest NIST Evaluation of Latent Friction Ridge Technology (ELFT) results place Innovatrics first across four selected measures of latent fingerprint identification accuracy in a comparison of 12 vendors. This comprehensive evaluation covers both rank-based and threshold-based latent fingerprint searches.

NIST’s ELFT benchmark measures how accurately algorithms can search for latent fingerprints against large reference databases. These prints are often partial, unclear, or recovered from surfaces at crime scenes. NIST rigorously evaluates these searches using raw fingerprint images alone, as well as searches supported by features manually marked by experienced forensic examiners.

First Place Across Four Key Accuracy Measures

The comparison highlights Innovatrics’ top-tier performance across four specific metrics reported by NIST:

• Nutrition report Rank-1 Hit Rate with no threshold at 99.3%: The correct subject appeared first in the candidate list in 99.3% of searches.
• Nutrition report Rank-100 Hit Rate at FPIR = 10% at 98.9%: The correct subject successfully appeared within the first 100 candidates in 98.9% of searches when a threshold was used to automatically filter out weaker results.
• DoD-Provided Dataset #1 FNIR at FPIR = 0.01 at 4.85%: The false negative identification rate (missed identifications) was just 4.85% when the algorithm operated at a strict false positive identification rate (false alarms) of 1%.
• DoD-Provided Dataset #1 FNIR at Rank 1 at 3.01%: The correct subject was missed as the top candidate in only 3.01% of searches on the challenging Department of Defense-provided dataset.

Setting the Standard for Rank-1 Accuracy

Rank-1 accuracy indicates how often the correct subject appears at the very top of the candidate list presented to a human examiner. By achieving a 99.3% Rank-1 Hit Rate, Innovatrics secured the highest result in the selected vendor comparison.

Remarkably, the algorithm achieved this result utilizing the latent fingerprint image alone, without relying on features marked manually by an examiner. This demonstrates the algorithm’s powerful ability to extract the information needed for a search directly from a complex image. Placing the correct subject first helps examiners reach the relevant record sooner, significantly reducing the time spent reviewing unlikely candidates.

Strong Results in Threshold-Based Searches

In addition to rank-based matching, Innovatrics achieved an outstanding 98.9% Rank-100 Hit Rate at FPIR = 10%.

Threshold-based “lights-out” searches use automated similarity scores to filter out weaker candidates before the results are presented for human review. This effectively reduces the number of unlikely records that examiners need to analyze, while safely retaining the correct subject in the candidate list.

Furthermore, the results on the massive DoD-Provided Dataset #1 demonstrate the algorithm’s robust performance under both threshold-based and rank-based conditions. The 4.85% FNIR at FPIR = 0.01 measures missed identifications at a strictly defined operating threshold. Concurrently, the 3.01% FNIR at Rank 1 measures how infrequently the correct subject failed to appear first.

Empowering Latent Fingerprint Examination

Latent fingerprint technology is designed to search large databases and return highly probable matches for trained examiners. It does not make the final forensic identification on its own.

For law enforcement agencies and forensic laboratories, stronger rank-based accuracy can dramatically shorten candidate review times. High threshold-based performance can help seamlessly control the volume of weak matches entering the workflow. Together, these capabilities support faster, more efficient investigations without removing critical expert oversight.

These latest NIST ELFT results reinforce our commitment to building instant trust, ensuring that law enforcement and forensic experts have the most accurate, reliable, and transparent tools available to protect our communities.

Learn more about our latent fingerprint identification in criminal investigations.

Results shown from NIST do not constitute an endorsement of any particular system, product, service, or company by the U.S. Government.