ROC scores 30% latent fingerprint biometrics accuracy gain in NIST’s ELFT

A new algorithm from ROC for matching latent fingerprint biometrics has scored a 30 percent accuracy gain in a U.S. government assessment.

The ROC +0014 algorithm was submitted to NIST’s Evaluation of Latent Fingerprint Technologies (ELFT) at the beginning of February.

The average false non-identification rate (FNIR) achieved by the algorithm across all datasets was 0.138, an improvement of nearly a third over the 0.198 scored by ROC +009, which ROC submitted to the ELFT in July.

ROC’s algorithm also made its matches in an average of 64 seconds, dramatically faster than the next entrant, which took 43 minutes. The company points out in an announcement that the faster matching time means less computing resources used and a lower total cost of ownership for investigative bodies.

ROC scored a rank-1 hit rate of 93.7, and a rank-100 hit rate at false positive identification rate (FPIR) of 10 percent of 88.8 percent.

“We are excited to once again be recognized as a leader in NIST ELFT testing. At ROC, accuracy and extreme scientific rigour are more than just priorities, they’re in our DNA,” says Dr. Brendan Klare, chief scientist and co-founder, ROC. “The same applies to innovation, namely enhancing operational performance while reducing system complexity and costs. Very few providers can match a 70 million fingerprint database in 1 minute.”

He notes that the production-ready algorithm evaluated by NIST is exactly the same as one as in the ROC SDK. The company makes the SDK readily available and free of cost for developers to test themselves.

ROC began submitting to the ELFT just a year ago, which the ROC SDK version 3 was introduced. The company’s latent fingerprint biometrics software was integrated into a specialized forensic camera at the end of 2024.

Article Topics

accuracy  |  biometric identification  |  biometric testing  |  biometrics  |  ELFT  |  fingerprint biometrics  |  NIST  |  ROC