Video: How Conversation Print Helps Fight Speech Spoofing

Learn more about biometrics and deepfakes at the next SpeechTEK conference.

Read the complete transcript of this clip:

Roanne Levitt: The spoofing problem is a big problem. So it started with patching together "my voice is my password" utterances by getting somebody to say all of those words individually and then putting them together and then using that to spoof a text-dependent system.

Well, there are technologies available that can detect playbacks and they can detect the acoustic signals essentially that are produced when you play something back through speakers. So anything that's played back through a speaker can be detected. So that's one aspect of it. The real challenge right now is all of the advances in technology for synthetic speech detection. So now you could, and it was mentioned this morning, Lyrebird is one example of a company who will allow you to go to their website and allow you to provide 30 utterances, 30 phrases, and then they will produce your synthetic voice, and then they could use that to try to spoof these systems.

But the technology's getting better and better in synthetic speech production and someday we'll be at the point where we'll have realtime voice conversion where you can actually have a fraudster talk into a microphone, or talk into their system, and have the system in realtime change their voice to your voice so they can actually live, answer the questions. So that's a real threat.

I know at Nuance we have furthered our research team that are continually looking into the synthetic speech production systems and detecting when the speech comes from these systems, they're detecting specific parameters, specific elements of what those systems are producing to be able to detect that those systems are actually producing that voice. But that's not enough because it's always a race. They're gonna one step ahead, we're one step behind, then we catch up, and then they move on step ahead. So we have to be very careful.

We've actually requested that the companies who produce these synthetic speech systems put watermarks in their transmissions and if they were to agree to do that then the nefarious folks would not be able to use it to try to spoof these voice biometric systems because we would be able to detect the watermarks, but unfortunately they've refused so other techniques are important.

One of the things that we've actually recently, a new innovation we've come up with, which will help a lot in this area, not just detecting synthetic speech, but it's a new biometric that we call Conversation Print.

So it's a conversational biometric. So now you're not only detecting how somebody produces sound, but you're also detecting how they choose the words they use, the grammar, their sentence structure, their vocabulary. If you think about it, if you were to speak to somebody you know well, or to a family member, and things would come out of your mouth which are not characteristic of you, they would probably notice that. So now we've actually been able to package that into technology. So many layers, you add all of these things together and again, it's a race.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

Video: Biometrics and Privacy Regulations Compliance

Nuance Communications Senior Manager, Commercial Security Strategy Roanne Levitt and ID R&D VP of Sales John Amein discuss the essential requirements for complying with privacy regulations in this clip from their panel at SpeechTEK 2019.

Video: How Biometrics Can Detect Deepfakes

Pindrop Director of Product Marketing Ben Cunningham discusses how liveness detection, synthetic speech detection, and other deep voice biometric technologies can fight the threat deepfakes pose in this clip from a panel at SpeechTEK 2019.

Video: Detecting Deception in Speech, Pt. 2: Findings

Interactions LLC NLP Scientist Yocheved Levitan outlines the findings of a recent academic study on detecting deceptive speech and how to distinguish it from truthful speech in this clip from her presentation at SpeechTEK 2019.

Video: Detecting Deception in Speech, Pt. 1: Methods

Interactions LLC NLP Scientist Yocheved Levitan discusses a recent research study on identifying cues of deception in speech, and machine learning-based approaches to classifying deceptive speech in this clip from her presentation at SpeechTEK 2019.

Video: What Is the Minimum Amount of Speech for Authentication?

Pindrop Director of Product Marketing Ben Cunningham discusses best practices for voice authentication in IVR design in this clip from his panel at SpeechTEK 2019.