Security Matters: Distinguishing Speaker Recognition from Speech Recognition

As more and more companies enter the “voice ID” space, from simple word spotting or speech recognition technology to the complexities of true voice biometrics or speaker recognition, much confusion has arisen with the former claiming to deliver what the latter does; in fact, they are quite different. How can a relying party, device manufacturer, online service provider, or mobile network operator understand the difference?


Separate and Not Equal

One incumbent technology, in existence for decades, is speech recognition. Many speech recognition companies are attempting to compete directly with voice biometrics providers despite the significant differences in the underlying technologies.

Speech recognition or word spotting allows systems to understand and respond to various voices and users who speak the recognizable word(s) in a particular language. Based on the phonetic properties of our speech the speech recognition system can transform into text any utterance spoken in the language for which the system is designed. For example, using speech recognition, a speaker could say, “wake up,” to prepare a mobile phone to respond. Optimized for low power and minimal security, word spotting based on speech recognition enables anyone who knows command words or phrases in the correct language to wake up a device and gain access to contacts, files and other data.

Voice authentication (also known as Voice Biometrics or Voice ID technology), such as AGNITIO’s, is a technology that is able to identify who is speaking. Based on the personal and unique characteristics that our vocal tract imprints in the audio waves, the system creates voiceprints that match to a person and recognize that individual to anything they utter. True robust Voice ID technology is language independent while speech recognition is language dependent and is also based on text dependency, making scalability harder to implement requiring high level of customization. AGNITIO Voice iD is language independent and works in various modes, ranging from text dependent to text independent where users have the option to set a passphrase or to speak in natural language to authenticate. Where voice authentication is employed, simply knowing a password or imitating a voice will not grant access to a device.

Additionally, any biometric technology must be protected for anti-spoofing where replay attacks through recordings are recognized, thus preventing access or authentication. AGNITIO Voice ID is protected for anti-spoofing detecting up to 97% of replay attacks. It is unclear how speech recognition and word spotting technologies can detect replay attacks as they are based on phonetics as opposed to a person’s vocal characteristics.

According to Biometrics Research Group, the speech recognition market is much larger and is expected to reach $20.1 billion in 2015. The voice recognition market, however, is much smaller, at $2.5 billion in 2015 according to the research firm – but growing much more rapidly, as companies strive to create a safer, more secure environment that protects their customers’ data and devices.

Clearly, there is more to secure voice biometric authentication than many vendors – especially those in the speech recognition space – currently purport.


Forming Alliances

As companies enter the market – or try to reposition their existing product line – partnerships with authentication technology providers are certainly expected to proliferate.

One such company that has played a prominent role in secure authentication – and with whom a number of companies have created strategic alliances – is Nok Nok Labs, a developer and implementer of the NNL S3 Authentication Suite.

Nok Nok Labs and AGNITIO are two of the six founding members of the FIDO Alliance, an open industry consortium dedicated to changing the nature of online authentication by developing technical specifications that define an open, scalable, interoperable set of mechanisms that reduce the reliance on passwords to authenticate users.

The FIDO Alliance now boasts more than 160 members and recently published the final FIDO 1.0 specifications for Universal Authentication Framework (UAF) and Universal 2nd Factor (U2F). With FIDO 1.0 specifications available, FIDO Alliance members can implement and broadly commercialize FIDO 1.0 specifications to make authentication simpler and stronger for all.

Upon the release of the FIDO 1.0 specs in December 2014, AGNITIO was the first FIDO Alliance member to offer a voice biometric FIDO Ready™ product. In early 2014, AGNITIO and Nok Nok Labs delivered a FIDO Ready solution demonstration winning the Envisioneering Design Award at CES 2014, showing how Voice iD, through Nok Nok Labs’ Multifactor Authentication Client, can perform voiceprint-based strong authentication in mobile applications.

This combination of Nok Nok Labs and AGNITIO Voice ID FIDO Ready™ products empowers users of online, mobile, and point-of-sale applications to conveniently access sites and services and perform transactions that are more secure, private and easier-to-use than passwords.

AGNITIO encourages additional biometrics companies to join the FIDO Alliance and to align with other member companies to implement stronger, standardized authentication in their products and services. The result will be a more secure and more convenient world for consumers and businesses alike.

Posted by Mike Goldgof, VP Marketing, AGNITIO
Post a Comment


Enter the characters shown in the image.