Verint Systems discusses identity authentication and fraud detection
Customer experience and engagement are crucial differentiators for a company in establishing a quality brand and generating repeat business. Providing fast and efficient customer service via call centers can often be a challenge for companies, as the sheer volume of calls and need to implement strong security measures can hinder the consumer experience.
Verint Systems Inc. has improved its voice biometrics-based identify authentication and fraud detection solution, providing a faster and more seamless approach to customer identity authentication, along with a safer and more secure sales/service environments.
Using Verint’s Identity Authentication and Fraud Detection, organizations can simplify the consumer phone experience when it comes to validating legitimate customers’ identities by easily and securely authenticating them based on their own unique voiceprints.
As part of its broader customer engagement optimization portfolio, Verint Identity Authentication and Fraud Detection can be easily enabled to provide immediate added security to existing Verint call recording environments, or be installed separately as a strategic advancement to existing engagement center environments.
In a phone interview with BiometricUpdate, Steve Williams, Verint VP and global practice leader of identity analytics discussed how the Identity Authentication and Fraud Detection solution works, why voice biometrics makes for a strong identifier, and how it compares to other biometric authentication methods.
Why did Verint decide to develop and release the Identity Authentication and Fraud Detection solution?
Steve Williams: It’s a dual-pronged strategy to both facilitate the consumer experience and to catch fraud more effectively. First, for the consumer, is to drive a frictionless authentication experience using that actionable intelligence from that live call. So we wanted to give consumers a better experience than they are currently getting when they go through this sometimes lengthy and arduous process of answering security questions at the beginning of each call. So the whole point is to use in place of that as much as possible biometric technology and match the voice that is being spoken on the line to a voice-print in the database and quickly move you through without having to ask very many of those questions or at all, so that you, as a consumer, can get right to the heart of the matter.
For the company that’s providing this technology, that handle time can cost them a lot of money. With voice biometrics, we can do that in a fraction of the time. So it’s a tremendous cost-savings to a contact center, not to mention the improvement in customer satisfaction by avoiding what is generally perceived to be an annoying process for the consumer. So we see it as a win-win for both parties, for the consumer experience and the company improve their brand image.
At the same time as we’re doing this to make a better experience on the consumer side, we’re also trying to reduce fraud by giving the company a better way to identify fraudsters. I’m sure well aware the consequences of the data breaches and all the information that we have out on ourselves on social media. Fraudsters are having the time of their lives getting through these security questions because they have the answers in front of them. Even though a fraudster can answer a knowledge-based authentication question, as well or better than consumers, We can catch them by printing their voice as well. So not only are we printing the consumer voice with their consent, but we are also printing any voices that result in a fraudulent activity on the account so that on every call that we hear that voice we can shut them down.
How does the Identity Authentication and Fraud Detection solution work?
There are two main flavors of voice biometrics in the market today. Active voice biometrics is text-dependent, which means it requires the consumer to enroll proactively and train the system to understand a catchphrase. At “such and such, blank, my voice is my password” would be a common way to do this. Well, in theory, the technology is fine for that purpose. But what we have found is that consumers don’t want to do this because it’s yet another point of friction in the process for them. So they haven’t adopted it. They’ve voted with their mouths, so to speak, and decided not to enroll actively in the system.
So consequently, we’ve been looking at this market for many years and determined that the better way to go about this is to do passive enrollment, which is a different flavor of voice biometrics. It’s based on mathematical modelling of the human vocal track and does not require a text-dependent catchphrase to be spoken and then understood on subsequent calls. It’s taking your physical characteristics, stores those, and then whatever that is represented in the form of the individual calling in, and they’ll just match a fingerprint, iris, or any other biometric. That enables the call center to not asking the customer to do more work. They just give consent that their voice can be used for security purposes, and then quietly in the background we passively collect the audio that will be used to make that voice-print.
That’s essentially how it works in the background of the live call. We’re taking the audio, and in some cases we can work off of historical recordings for that customer, and create a voice-print that will be used on subsequent phone calls. The same is true for fraudsters, of course, although we don’t need their consent to enroll their voices in the “bad” database.
What makes voice biometrics a good identifier and authentication method for the financial industry?
It’s very accurate. We’ve gone through multiple generations of voice biometric solutions in the past 10 to 15 years. So it’s an extremely accurate methodology of identification. And it’s the most appropriate biometric to use for those that are calling into a contact center. There are going to be a proliferation of biometrics, such as using your fingerprint to buy something on your iPhone through Apple Pay. But when you’re calling in to a contact center, the most common modality is going to be your voice. It’s highly accurate and it’s unique to each customer.
Do you think it is stronger than other available biometric authentication methods, such as fingerprint or iris authentication?
I haven’t done the research the accuracy level by comparison. All I can tell you is that we’re in the high 90 percentile of accuracy which I think for any biometric solution is pretty comparable.
Many other identity authentication solutions pair voice biometrics with a second or even third biometric identifier to create a two or three-factor authentication method. Is this something that Verint intends on offering in the future?
We believe that one factor alone won’t work in every circumstance, despite its accuracy. You’re just going to have situations where much like any other biometric, there’s always going to be some percentage of cases where that’s not going to answer the question fully. Fraudsters have certainly proven to be very clever and demonstrated the ability to evolve with the technology. In many cases, stay a step ahead of commercial efforts. What we feel is the better way to do this is to combine voice biometrics with sophisticated data science capabilities.
We have a meta-data analysis, a predictive analytic solution that goes through and looks at a variety of data factors that surrounds the call. This is additional actionable intelligence that we’re extracting from the call itself. Things like, how many times have we seen this particular phone number calling into our contact center and across different accounts or not. So there’s a lot of information just on the basis as simple as the phone number. And there’s some additional data from the network that lets us to gather more intelligence as to its authenticity. And then there are a huge set of behavioral factors that we also assess that helps determine the likelihood of fraud.
In this first release we just have the voice biometric component, but we will be adding the meta-data analysis along with that in the future.