Voice biometrics technologies are finding a comfortable and growing niche in an opportunity area that intersects with rules-based systems, speech analytics, deep neural networking and other flavors of “artificial intelligence.” With the first half of 2018 behind us, it is time to take stock of the technological developments and regulatory factors that shape both demand and deployments and document the trends that are fueling growth and giving comfort to both developers and deployers.

Global Survey of First Half Developments
The first five months of the year were infused with concern over Europe’s General Data Protection Regulation (GDPR), implemented on May 25 ( or “#Privmas Day” to some). In spite of concern over enrollment methods and whether a voice print is “personally identifiable information” (PII) voice biometrics has continued to grow in adoption. As an example Nuance Communications announced that it reached the 300 million enrolled voiceprints, and 5 billion annual authentications milestone.

Nuance’s milestones were reached amid news from a number of vendors showcasing Artificial Intelligence (AI) and Machine Learning (ML) and a variety of multi-factor and omni channel use-cases. These included “voice-enabled cognitive rooms” that result from a partnership between Harman and IBM Watson. First implementations are expected to support health care and hospitality, where it will quickly become important to use a flavor of speaker identification to extend personalized and secure services to patients or hotel guests.

Global uptake was also evidenced in Singapore where over 56% of firms have either implemented or are implementing voice biometrics-based solutions.

Here is a quick update for those who need to catch up on the goings-on in the world of Intelligent Authentication, Voice-Based Authentication and Fraud Management:

Interpol announces Speaker Identification Project
The blogosphere lit up when Interpol launched Speaker Identification Integrated Project (‘SiiP’), which is a fusion of speech analytic algorithms, including speaker recognition, through the matching of speech samples taken from live phone calls or social media posts to databases of known criminals’ voice recordings in databases shared by law enforcement and other public security agencies.

HMRC Voice ID: Implied versus Explicit Consent
Hot on the heals of #Privmas, UK’s taxation body, Her Majesty’s Revenue and Customs (‘HMRC’), came under the attack from privacy campaigning organization ‘Big Brother Watch’, for failing to obtain explicit consent from the circa 5.1 million taxpayers whom it claims were “railroaded” into registering for the Voice ID service. While the security and operational efficiencies of Voice Biometrics stand strong, the emphasis on explicit consent is driving a redesign of the user permissioning processes.

AI, DNN, Machine Learning-Infused Solutions Support Mobile- and Voice-First
As mobile-first and voice-first drive modern UX design strategies, the variety of sensors on smart phones is spawning a multitude of biometric permutations using Voice Biometrics for stronger, step-up and out-of-band authentication. Together with Deep Neural Networking (DNN) and Convolutional Neural Networking (CNN) technologies, voice biometrics has benefitted through massive performance improvements as well as adjacent modalities that are transitioning from academia into the commercial world. Cases in point:

• Nuance added ConversationPrint, to its product line. It is a ‘behavio-conversational’ innovation that can identify fraudulent activity in real-time based on choice of words, and patterns of speech or writing, during a speech interaction with a human or virtual assistant; together with their Gen3 DNN voice biometrics engine.
• Sensory’s TrulySecure SDK-based platform uses AI-based dynamic modeling to passively identify enrolled users, and enrol previously un-registered users. Sensory also partners with prominent FIDO-certified solution providers HYPR and Nok Nok Labs.
• Pindrop launched Deep Voice, based on ‘end-to-end’ DNN’s that deliver greatly improved performance, while also being able to identify virtualized activity such as ‘bots’ and synthesized audio; together with Toneprinting as part of their Passport portfolio.
• Phonexia’s Deep Embeddings DNN-only powered voice biometrics engine claimed to have achieved a greater than 100% improvements in speed and accuracy
• ID R&D has achieved a 40% improvement in authentication accuracy, together with reduced software footprint of down to 2MB for on-device applications, and is now available on the Samsung SDS Digital Identity Platform.
• InnoCapital the corporate venture vehicle of Banco Sabadell’s InnoCell digital hub has led a EUR1 million investment round in voice authentication startup Biometric Vox.
• Interactions was awarded with ‘Best Use of AI for Voice Biometrics’ by the Artificial Intelligence (AI) Breakthrough Awards.
• Another concentration point for AI is the automobile, where biometrics is playing a increasing role in speaker identification and authentication for security use-cases like vehicle unlocking, emergency and roadside assistance and imminently payments as well. For example, Nuance’s Dragon Drive adds voice biometrics and advanced audio processing technologies to recognize and allow multiple passengers to interact simultaneously with the automotive assistant.

Multi-factor, Multi-modal, Centralized, Hybrid and Distributed Systems
As mobile progresses towards the priority channel, decentralized architectures are gaining greater favor due to improvements in device-based, “small footprint” voice biometric engines, as well as the benefit of distributing the voiceprints and any associated personal identifiable information (PII). As mentioned above, device-based systems are able to utilize the multitude of sensors available to orchestrate a variety of multi-factor authentication (MFA) combinations for risk-based, step-up and other authentication strategies. For example:

• FIDO’s rapid progress and global expansion is a strong indicator of the preference for distributed solutions
• Similarly, REL-ID from Uniken also binds Biometric + Certificate + Hardware Token of the device
• Daon in collaboration with Digi-Sign, a subsidiary of Hong Kong e-commerce services provider Tradelink announced a successful implementation of it’s multi-factor, FIDO compliant IdentityX platform in the region.

Cloud based systems are also gaining traction due to the ease of implementation and cross-channel convenience:

• Aimbrain’s ‘On Demand’ provides app developers self-service SDK’s to swap PINS, passwords and codes for facial and voice user biometric verifications. The company is also disrupting commercial consumption-based models through its free-for-life 1,000 API calls/month offering to small users or enterprizes wishing to try-before-they-buy.
• Aware has added voice and behavioral (keystroke) biometrics to their cloud-based fingerprint, face and iris offerings.

Product, Solution, Regional and Vertical and Use-Case Innovations
Various product enhancements and use-case applications were also announced:

• Joining its international cousins, ING bank in Australia rolled out voice activated banking with Siri
• Chennai-based WFO specialist Zuper launched a voice biometrics-based remote timesheet management applications for field and remote workforce.
• Nuance earned a Silver Stevie Award for the Best Use of Technology in Customer Service through use of voice biometrics to strengthen BBVA Bancomer’s pension system Proof of Life verification.
• While not new, Voice Signatures for contract signing is gaining momentum in Spain and Portugal
• Rostelecom announced the launch of the beta version of its Unified Biometrics System (UBS) in partnership with three Russian banks to integrate voice and facial recognition into their systems.
• Fortr3ss applies Zero-Knowledge Encryption to its Password Manager to ensure that user passwords, notes and other confidential data is only accessible to the user, accessible by their voice.
• Startup BABB (Bank Account Based Blockchain) raises $20 million in ICO for biometrics and blockchain-based banking platform. BABB has patents pending for technology related to face and voice biometrics, as well as its decentralized payment card.
• The BISON (BIg Speech data analytics for cONtact centers) project has used a portfolio of speech technologies, including both speech analytics technologies (speech transcription, keyword spotting) and voice biometrics in multiple European languages to uncover cover diverse use cases in contact centers.
• Česká spořitelna, the Czech Republic’s largest bank, is also the first in the country to use Nuance’s text-independent (FreeSpeech) voice biometrics to enable customers to simply speak in a natural conversation with a contact center agent order to authenticate themselves.

These developments and product introductions reflect an acceleration in the adoption of solutions for authentication, identification and fraud reduction driven by greater capabilities and reliability. Opus Research expects the momentum to continue to build during the second half of the year.

Categories: Conversational Intelligence, Intelligent Authentication, Articles

Tags: PhonePrint, Aware, Voice biometrics, Deep Voice, Zuper, Nuance, phonexia, Voice Signature, Sensory, Deep Embeddings, Fortr3ss, ID R&D, Biometric Vox, HMRC, FIDO, explicit consent, REL-ID, VoiceFirst, Daon, ConversationPrint, Identity-X, Pindrop, Aimbrain