ID R&D Takes Cue From Neural Network Training in Fight Against Deepfakes

ID R&D has been granted a new US patent for a novel approach to the detection of presentation attacks against a facial recognition system. And the company says its method has already improved the speed and accuracy of its IDLive Face anti-spoofing solution, designed to eliminate the kind of friction that anti-spoofing systems sometimes deliver for end users.

ID R&D’s patented system takes a cue from innovative approaches to the development of artificial intelligence that use visual transformers to focus AI algorithms on the most important data during neural network training. ID R&D’s solution likewise trains a liveness detection algorithm to focus only on the most relevant parts of a face image when scanning for signs of a presentation attack.

Focusing only on the most salient data points of a face image not only makes the liveness detection system faster, but also improves its accuracy, the company says.

In announcing the patent win, CEO Alexey Khitrov highlighted the threat of AI-generated deepfakes, which prompt the need for anti-spoofing technologies that are sophisticated enough to detect them.

“Rapid advancements in artificial intelligence bring new challenges to securing digital onboarding and authentication — particularly from deepfakes — and we expect this trend to persist,” he said. “Fortunately, we are also working hard to research and develop new ways to leverage AI to its full potential to counter AI-powered fraud.”

ID R&D, which operates under the umbrella of parent company Mitek, has previously demonstrated the value of its investments in technological R&D. Last year, for example, the company touted the results of independent testing that found its IDLive Face solution showed virtually no bias across demographic groups like race, gender, and age — a matter of growing concern in recent years as the use of facial recognition technology has become more widespread.

ID R&D’s technological portfolio extends beyond facial recognition, however. News of the company’s latest patent approval comes after the company brought voice recognition to TECH5’s multimodal T5-Digital ID platform toward the end of last year.

–

June 21, 2023 – by Alex Perala