ROC.ai upgrades PAD capabilities of face biometrics SDK

Rank One Computing has updated its software development kit with what the company says are substantial improvements to its biometric presentation attack detection (PAD) performance.

The ROC SDK v2.7 delivers significantly higher accuracy than previous versions, according to the announcement.

ROC.ai recently passed a Level 2 PAD assessment from iBeta. The company also submitted the PAD algorithm from its SDK v2.4 for NIST’s PAD evaluation earlier this year, in which the company scored the best results for both convenience and security for detecting evasion attempts with presentation attack type 6.

From the v2.4 algorithm to the v2.7 algorithm, however, ROC.ai claims it has dramatically reduced the PAD error rate. The company has found its latest PAD algorithms deliver an attack presentation classification error rate (APCER) of 0.022 at a bona fide presentation classification error rate (BPCER) of 0.01. The represents a by 5.5 times-gain for convenience. Similarly, at an APCER of 0.01, ROC.ai delivers a BPCER of 0.021, a quarter of what it was for v2.4.

The company also highlights the results of its technology in the NIST test for video replay impersonation attacks (0.9 percent error rate), and for detecting masks.

ROC.ai has just appointed a VP of business development to grow its customer base.

Article Topics

biometric liveness detection  |  biometrics  |  face biometrics  |  presentation attack detection  |  ROC  |  SDK