Automating the Edge with Robotics

ID R&D: edge computing opens up faster, more secure experiences through voice biometrics

Categories Brand Focus  |  Edge Computing News
ID R&D: edge computing opens up faster, more secure experiences through voice biometrics

ID R&D is a developer of AI-based voice and face biometric software components that simplify the process of authentication while providing the strongest protection against sophisticated identity theft and fraud techniques.

With the adoption of voice interfaces in an increasing array of consumer devices and services, the executives say the company’s goal is to provide consumers a frictionless experience that doesn’t sacrifice security—and the power of edge computing is an important part of that mission. To achieve this goal, ID R&D established a strong position as a technological leader in voice biometric space with multiple first place finishes in the leading independent benchmark evaluations. A recently awarded patent for Text Dependent Speaker Recognition solidified the company’s standing.

In June 2021, Mitek Systems acquired ID R&D for $49 million to integrate its face and voice biometrics and liveness detection capabilities. By integrating ID R&D’s biometric technologies, Mitek says it can simplify and secure the entire transaction lifecycle for businesses and consumers with a single authentication solution. ID R&D continues to be operated as an independent division.

Alexey Khitrov, CEO at ID R&D, talked to Edge IR about ID R&D’s role as a provider of security and authentication technology and his perspective on edge computing.

Edge IR: Why does edge computing matter for your company, and where does it matter?

Alexey Khitrov (AK): As you know, edge computing is about pushing more computing power closer to the point of user interaction for the goal of creating a faster, better user experience. We see edge computing as a way to open up new experiences that are challenging to implement in the cloud.

For example, people are now comfortable talking to smart speakers in their homes. They’ve learned to use smart devices in a variety of ways. Now imagine that you want these devices to know who you are so they respond differently to you compared to your spouse. Or maybe you want to limit who can use the device, like parental controls, to prevent your child from ordering a crate of gummy bears. The technology to know who is talking is voice biometric technology from companies like ID R&D. Yes, the voice biometrics could be run in the cloud, but in terms of speed and efficiency of processing, pushing this capability to the edge is far better for the user.

We’re excited because as edge computing processing power is increasing, at the same time our investment in R&D is leading to faster, smaller, and more accurate voice biometric algorithms. This combination of more powerful edge devices and more efficient algorithms will lead to significant deployments of voice biometrics running on the edge in 2022.

How are you working to manage the complexity of models and computational power that’s available on edge devices?

AK: Some of the innovations we’re bringing to market are not only about enabling core biometric capabilities at the edge, but the ability to minimize the footprint of these technologies and successfully run them with the limited resources available. We’re one of the first companies to put voice biometrics on a neural processing unit (NPU) in order to provide security and authentication in a system-on-a-chip (SoC) deployment.

For instance, we have integrated our voice biometrics and voice anti-spoofing onto the NPU of the Synaptics VS600 high-performance multimedia SOC solution. The ability to tap into the NPU enables the use of sophisticated algorithms that result in highly accurate AI-based voice biometrics on an extremely small footprint. The solution works with an array of voice-enabled smart home devices including set-top boxes, smart speakers, displays, and security systems. For instance, we have the solution working on a voice-enabled TV remote control. Not only does it enable instant identification of the speaker for personalization, but it also provides security for parental controls, subscription changes, and payments. Imagine selecting a new movie release and authorizing payment in one simple command — with no passcodes or PINs. We can even confirm that the voice is not somebody who has just recorded your voice.

In some ways, smartphones have led and will continue to lead in terms of technology on the device, including AI processing and LIDAR on the iPhone, for example. What’s the opportunity for your company in mobile?

AK: Mobile is one of the key segments for us. We work with mobile manufacturers to deliver voice biometrics to their devices. The reason is simple — there are many circumstances where mobile users need to be handsfree or otherwise want the convenience of commanding their mobile device via voice, but they don’t want the mobile device to work for anyone else. Voice biometrics makes that possible — the phone will know and respond to your voice but no one else’s.

Voice biometrics have been around for a while; however, what’s new is that until our latest generation of algorithms, the ability to use voice to command the device did not meet security standards like those associated with Android. In other words, it would too easily hear someone else talking and think it was you. This number is in the range of 1 in 50 people could probably match. This is called the impostor accept rate. We have developed and are now releasing voice biometric algorithms so that the impostor accept rate is reduced to 1 in 50,000. Now we’re at a number that security people respect and that meets industry standards for security. And we are achieving these accuracy numbers while making the solution small enough to be deployed right on the Edge. This is a huge breakthrough, not just for the company but for the entire industry!

What are some of the challenges of voice biometric matching, and how are you addressing those challenges?

AK: Today, synthetic voices or “deepfakes” can be created using a limited amount of audio. While there are a lot of good uses for this technology, we’re also seeing it used for fraud. For example, the story in which an executive was tricked into transferring money to a criminals’ account based on a call that he thought was from his boss. Just matching voices is not enough — it’s critical to determine if the voice is live or a spoof. The same goes with facial recognition. It’s not just matching people, it’s being able to tell the difference between a person and a representation of that person. We provide voice and facial liveness capabilities [to do that].

Technology is still a differentiator in this market, and it’s still about delivering capabilities that we were not able to before to enable transactions through new channels. The industry really is on the edge of rapid growth.

Article Topics

 |   |   |   |   |   |   | 


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Featured Edge Computing Company

Edge Ecosystem Videos

Automating the Edge


Deploying AI Models at the Edge


Latest News