JOHANNESBURG, SOUTH AFRICA (REUTERS – A groundbreaking technology is promising to revolutionise electronic identification and authentication, according to its South African developers.
aiThenticate Computervision Labs, working with the University of Johannesburg, say aiDX represents a quantum leap over conventional facial recognition systems.
Nearly seven years in development, the system uses artificial intelligence (AI) to simulate human cognition, going beyond the biometrics of passwords, fingerprints and eye scans, and can be used in larger populations with more accuracy, its developers say.
“It came from the initial idea that conventional methods of authentication really fall short in the post 9/11 world, where we need to have a much greater level of a certainty who somebody actually is. In order to answer that question, we appealed to deep science, deep learning, to develop an artificial intelligence method to replicate or to mimic or to simulate the way that we as humans, intuitively and instinctively go by recognizing somebody’s head. It is very different to the conventional traditional way of face recognition, finger print recognition,” said aiThenticate CEO André Immelman.
The process works in the same way the human sensory system is able to identify and authenticate information.
The role of the eye is played by a camera which feeds information using a cable, similar to the retina, into the brain, or in this case the cloud where the information is processed.
Unlike conventional biometrics, such as fingerprints, faceprints, voiceprints and irisprints, which are all based on simple geometry: connecting key features to form a pattern that is then associated with a particular individual, aiThenticate Computervision Labs’ technology uses deep science to answer the “who is someone actually?” question, says Immelman.
He says the technology is simple enough to use on a mobile phone but so intelligent it can benefit high level, big-data entities like governments, big corporations and even law enforcement.
“The technology is not intended or designed as a light weight application for anybody to unlock a phone or something like that. It is designated for serious real world commercial applications, conventional methods failed very quickly as the sample size increases, is useful for locking your phone but for very large applications, serious applications, simply doesn’t work. So typically it has applications in the security sense, it has applications in a customer services sense, you know this kind of things the bank calls you up and says: this is your bank calling, please, where you live, what is your mother’s name, what’s your dog’s favourite hobby, whatever the case it may be. It take that kind of guess work out of the equation completely and it answers the, “who” question to a much greater level of confidence or certainty, than what traditional or conventional biometrics have been able to do in the past,” he said.
More than 15 million consumers were victims of identity theft last year, losing 16 billion US dollars according to a report by Javelin Strategy Research, which tracks cases of fraud in the U.S.
aiThenticate says up to 2 trillion US dollars was lost globally to identity theft.
“To be effective at this performance of our technology, we need to continually make sure that we get people to use it, therefore we need world attraction how to use it. So the more users or the more subjects use the technology, the more the effectiveness, the robustness and the capability that carries on,” said Dr Olaleken Komnsane, an engineer at aiThenticate labs.
The developers say field tests so far have shown that aiDX capabilities surpass the accuracy of conventional methods by up to 20 times.
Plans are underway to take aiDX technology to the market in the next month.