ID R&D removing the user experience but keeping the biometric security in authentication
As digital identity proving requirements have proliferated, and biometric technology has become part of the regular routines of millions of consumers, some of the limitations of traditional approaches to authentication have been exposed. Facial recognition often asks for additional, sometimes awkward actions from users, such as blinking and smiling, in order to guard against spoofing. Authentication in call centers likewise involve interactions over a significant length of time for text independent voice biometrics, and provide customers with a user experience in some ways hardly better than passwords.
ID R&D is developing biometric technologies that minimize the user experience, or eliminate it altogether, without compromising on security. By running biometric authentication continuously and passively in the background, the security of transactions is increased without impacting user experience.
This enables new kinds of secure interactions, and has only become possible due to new artificial intelligence techniques, Chief Science Officer and co-founder Konstantin Simonchik told Biometric Update in an interview.
“In voice biometric technology, the so-called ‘i-vector’ approach has generally been used for the last three or four years,” Simonchik explains. “i-vector estimation is based on the generative training approach, and it works relatively well with small amounts of data. But now we’re using a convolutional neural network to extract what we call the ‘x-vector.’”
The difference is that with a larger amount of data, the x-vector provides more accurate matching, and does so with less authentication input. It made sense to use the i-vector when the technology was in early development, and a limited amount of training data was available, Simonchik says.
“We invested significant resources in building one of the largest voice biometric data sets, this allows ID R&D to further its development capabilities,” he adds.
“It really brings a new level of accuracy, and requirements for the lengths of audio, which is incredibly important for conversational interface where sometimes conversations are fairly short,” CEO and co-founder Alexey Khitrov tells Biometric Update. “Comparing to i-vector technology that takes usually hundreds of Mbytes footprint size and RAM x-vector approach is ten times smaller, this allows to use x-vector methodology for wide specter of mobile devices”
When ID R&D launched a new version of its text-independent voice biometric verification product IDVoice, it declared it the fastest and most accurate technology of its kind.
Another limitation of traditional biometric approaches is that anti-spoofing techniques have been challenged by conditions that affect the quality of input, including environmental factors such as lighting conditions for facial recognition, Simonchik says. Companies have typically used convolutional neural networks (CNNs) to resolve these challenges, but ID R&D has developed an ensemble of deep neural networks and CNNs using boosting methods.
This fusion performs anti-spoofing checks within a single second with a very high degree of certainty.
“This completely changes the user experience paradigm, because now the authentication happens without user experience,” according to Simonchik. “It’s been moved into the background, using multiple biometric layers and all of the information we can collect.”
ID R&D was founded only two years ago, but has brought to market IDBehave, IDVoice, ID Anti-Spoof, the flagship mobile login solutions IDSquared, which combines facial recognition and behavioral biometrics. It also launched SafeChat, which the company says is the first completely frictionless continuous multi-modal biometric verification product for chatbots and virtual assistants, late last year.
“These technological advances are just arriving on the market,” Simonchik says. “SafeChat is a unique solution that provides a fusion of modalities, and by combining voice with facial recognition and behavioral recognition, we have built a completely frictionless biometric experience for a growing range of applications. Interactions with chatbot and virtual assistants are just the beginning.”
In addition to smartphones and the channels where voice has been used in the past, the company is also developing its technology for digital assistants and IoT integrations, such as with smart home products, smart appliances and automobiles.
The company is participating in the Fintech Batch of Plug and Play’s accelerator program, as well as the Google Cloud Startup Program, giving it additional resources to further drive its technological innovation. It has also been recognized by Microsoft’s Global Startup competition as one of the Top-10 AI startups in America, and selected by TechCrunch’s Disrupt SF as a Top Pick for its fintech products.
“Every step of the way we’re looking at multiple modalities combined and fused together to provide the enterprise with this level of authentication while there’s no user experience for the actual end user,” Simonchik says. “Voice is getting out of the call centers, into mobile and into the wider conversational interface.”