CyberLink has cleared one other requirements milestone. The corporate’s FaceMe facial recognition engine has handed the second degree of iBeta’s Presentation Assault Detection (PAD) take a look at, after beforehand clearing the primary degree of the take a look at in November of 2021.
The accomplishment attests to FaceMe’s skill to protect in opposition to completely different types of spoofing. In that regard, the Degree 1 PAD take a look at evaluates an answer’s skill to detect the usage of 2D movies and pictures. The Degree 2 take a look at, however, seems at extra refined 3D masks, whether or not they’re printed or made with lifelike supplies like resin or latex. FaceMe acquired an ideal rating in each evaluations, and was efficiently capable of detect each single one of many spoofing makes an attempt that was made in opposition to the platform.
In additional technical phrases, that implies that the corporate achieved an Assault Presentation Classification Error Fee (APCER) of zero p.c. FaceMe was additionally good at distinguishing a registered person from a human imposter, with Bona Fide Presentation Classification Error Charges (BPCER) of 1.5 p.c on iOS gadgets and a couple of.5 p.c on Android. These outcomes point out that false matches are extraordinarily uncommon, and that FaceMe is subsequently a dependable type of authentication for on-line banking and different digital functions.
“Biometric fraud prevention is likely one of the most crucial imperatives for sectors of the financial system resembling fintech, the place a safety breach can take catastrophic proportions,” mentioned CyberLink CEO Jau Huang. “The iBeta PAD take a look at is widely known as the worldwide commonplace for the spoofing prevention efficiency of facial recognition choices. Reaching Degree 2 with an ideal rating anchors FaceMe as one of many highest options to guard entry from biometric fraud.”
Options that cross the iBeta take a look at are compliant with the ISO-IEC 30107-3 commonplace. The group itself is accredited to check facial recognition options below the NIST’s Nationwide Voluntary Laboratory Accreditation Program (NVLAP).
CyberLink additionally posted a top-three rating in the latest ICCV anti-spoofing problem.
–
March 22, 2022 – by Eric Weiss