The applications of facial biometrics across many sectors eliminates any doubt as to the accuracy of artificial neural networks to match live facial images of people with advanced data.

While this fact is not in dispute, researchers at Kyushu University have recently expressed the view that AI which powers image recognition can fail to identify a person if an image is modified.

Through coming to understand the shortfalls of AI-powered biometrics, researchers have published a method called ‘Raw Zero-Shot’ that assesses how neural networks work with modified elements of a person’s appearance or photo. The aim is to understand all the potential features that may not be compatible with current detection capabilities of the AI in order to make systems more robust.

Explaining this concept, Danilo Vasconcellos Vargas who led the study said: “There is a range of real-world applications for image recognition neural networks, including self-driving cars and diagnostic tools in healthcare”.

“However, no matter how well trained the AI, it can fail with even a slight change in an image.”

In all applications where digitalisation and AI replaces a manual process, technology can not be guaranteed to always outperform human judgement in detecting errors or discrepancies. Facial biometrics can and have faltered whilst being deployed on such a scale to replace manual procedures such as border control around the world.

In fact, experts have found that while an doctored image may appear unchanged to the human eye, AI will not be able to accurately identity it. The team tested numerous image recognition AIs against photograph samples they had not been trained to detect in order to establish behavioural patterns.

Other global leaders in the field of biometric authentication are also seeking to significantly improve the accuracy of deep-learning algorithms that power facial recognition technology to secure safe international travel as well as streamline operations in the private sector.

The NEC was recently awarded a commendation by Japan’s Minister of Education, Culture, Sports, Science, and Technology in April to recognise the high-accuracy of its facial recognition technology in contributing to safety and security in the aviation sector.