By Roy Urrico
Credit unions, like all financial institutions and enterprises that conduct business online, must tread a fine line in authenticating users while not interfering with the member experience. Nuance Gatekeeper, a cloud-based biometric security solution, allows authentication while detecting and stopping fraudsters wherever — and however — they engage. This includes communication via voice or text.
Burlington, Mass.-based Nuance Communications, a technology pioneer in conversational artificial intelligence (AI) and ambient intelligence, features Gatekeeper as part of the Microsoft Digital Contact Center Platform, an open, extensible, and collaborative platform for omnichannel customer engagement. (Microsoft closed on its approximately $16 billion acquisition of Nuance in March 2022.)
Nuance Gatekeeper provides both authentication and fraud prevention on a single platform that spans voice and digital channels. Gatekeeper analyzes how a person sounds, talks or types, and how they behave, while checking their device, network, location, and other factors for signs of fraud, explained
, vice-president and general manager of the security and biometrics line of business at Nuance. All of this happens instantly in the background of each engagement, so customers get the service they need without remembering passcodes, PINs, or security questions, Beranek added.
Nuance confirmed credit unions deploying Gatekeeper so far include the $5.1 billion North Chesterfield, Va.-based Virginia Credit Union (VACU) and the $7.9 billion Apple Valley, Minn.-based Wings Financial Credit Union, which was a runner up in the 2022 Finopotamus Tekkie Awards for deploying Nuance’s Intelligent Engagement Platform.
Cool but Efficient Technology
Fraud threats continue to increase and occupy credit unions’ cybersecurity strategy. “The fraud community is attacking these smaller organizations where a couple of years back their focus was elsewhere," said Beranek. “We've seen a groundswell of demand for our biometric solutions with credit unions.”
Nuance, an innovator in voice biometrics through its Dragon Speech Recognition Solutions, took its accumulated knowledge of voice communication to thwart fraudsters in a new way. One of Nuance’s “cool innovations,” stated Beranek, “is applying our technology, not from a voice perspective, but from a text perspective with our conversation print technology.”
Beranek added, “What we learned through voice biometrics we are applying to language. Our natural language understanding was instrumental in creating these (conversation print technology) models to identify people.”
Gatekeeper can analyze vocabulary, grammar, and sentence structure as well as punctuation and emoji use to determine if agents have a fraudster in front of them instead of the legitimate member, Beranek noted. “Our roots in Dragon definitely came in handy with the Gatekeeper solution.”
The process is not to be confused with keystroke dynamics or typing biometrics, which refer to the automated method of identifying or confirming the identity of an individual based on the manner and the rhythm of keyboard input.
“That's a different technology,” noted Beranek, “We are solely looking at the vocabulary, grammar, sentence structure and punctuation. We are using those data inputs to identify if this is fraudulent or not. You can think about conversation print as another layer of security that a credit union can use to secure their members.”
Gatekeeper’s Security Fence
Beranek recalled conceptually that Nuance had this technology in its “bag of tools” for a while, but decided to roll it out for overall accessibility as the volume of fraud attacks that occur in text-based interactions increased. “That surprised me just as much as I think it surprised our customers. There was this basic assumption that we all had in the industry that fraudsters preferred the phone channel, where they could socially engineer a contact center agent.”
Beranek pointed out that cybercriminals try to take advantage of service representatives’ willingness to help members. “The prevalence of fraud attacks in text-based interactions is higher (than first thought). We found with one organization that it was twice as prevalent in the live chat channel as it was in in the phone channel. We have realized that the reason why the fraudsters are comfortable with this (chat) channel is that it gives them an extra layer of anonymity.”
Beranek also noted that credit unions increasingly use the conversational AI piece of Nuance technology, weaving in biometrics within member engagement. “At the same time while we're automating those experiences with artificial intelligence, we're also making sure that those experiences are secure and that fraudsters aren't taking advantage of those automated experiences to compromise accounts.”
Roanne Levitt, senior director of strategy and innovation at Nuance Communications, told Finopotamus at the recent FinovateFall 2022 conference held in New York City, “We have been focusing quite a bit on making that experience very painless, frictionless, not only for the customer, but also for the agents and to automate and provide intelligent systems within the whole flow to help customers get what they want done.”
Looking Behind the Technology
Beranek said, “My hypothesis when we first developed this technology was it would work great if somebody was typing an email and you had a sufficient amount of text. But it is surprisingly accurate with the types of interactions you would have with live chat.” This means Gatekeeper’s conversation print model works with just a typed sentence or two.
Levitt added, “Voice is the foundation, what we are known best for is our incredible voice technology. But in the digital space, oftentimes we do not have a voice to use.”
Levitt noted an engagement could be a chat session, “Where we do not have that voice, but we do have what the customer is typing. We all have kind of a style in how we chat and our emoji use, for example, our use of words and sentences and tenses, all these things. So, in our business unit, we look at everything to see if we can recognize an individual, but also recognize patterns in and how fraudsters and fraud rings operate.” This might also include tenses and scripts.
If during a live chat session Gatekeeper detects the representative speaking to an unauthenticated individual, the Nuance process recommends the agent continue the live chat interaction. “Just so that we get as much data as possible, but we freeze the account, we freeze the transaction, so it actually doesn't go through,” pointed out Beranek.
“I'm really proud of our research team, it's always hard to wrap your mind around how we can identify somebody even with very little data,” Beranek said.