ParallelChain Lab to unveil AI at Interspeech 2024 in Greece
ParallelChain Lab, a Singapore-based software company specialising in digital identity-related artificial intelligence, is set to present its voice anti-spoofing AI at the Interspeech 2024 Conference in Kos Island, Greece.
Interspeech is the world's largest conference on speech processing, organised by the International Speech Communication Association (ISCA) and sponsored by Meta, Apple, Google, and Amazon.
This year's Interspeech includes the 'ASVspoof5' challenge, which has garnered global attention for developing AI systems capable of detecting AI-generated voice deepfakes. An expert panel comprising representatives from Microsoft, the National Institute of Informatics (Japan), EURECOM (France), and Carnegie Mellon University evaluated nearly 100 entries. ParallelChain Lab's entry was among the few selected. The company's Vice President of Engineering, Panagiotis Simatis, will present this breakthrough technology.
ParallelChain Lab's system not only excels in detecting AI-generated deepfakes and presentation attacks but also offers advanced voice-matching capabilities. This enhances the protection and accuracy of voice-based verification systems by providing robust two-factor security.
While facial recognition serves as the standard for biometric-based identity verification today, voice processing offers notable advantages in both security and privacy. The ubiquity of photos on social media can provide malicious actors with enough material to create convincing facial deepfakes. Conversely, voice biometrics are more difficult to replicate and thus are less susceptible to manipulation.
Additionally, voice biometrics are inherently less invasive, making them ideal for sensitive contexts. For instance, age verification protocols on adult platforms can benefit from voice biometric systems, which avoid the discomfort and privacy concerns associated with facial scans. These systems also alleviate substantial regulatory risks for social media platforms managing the facial data of underage users, which have previously led to significant fines.
With the increasing integration of smart home and Internet of Things (IoT) applications into daily life, advancements in voice recognition are critical to prevent unauthorised access via spoofing attacks such as audio playback and deepfakes.
ParallelChain Lab has been pioneering in voice-processing AI. Earlier this year, the company developed a voice-based age estimation AI for the Personal Data Protection Commission of Singapore. The authority sought privacy-enhancing alternatives to face-based age verification methods.
ParallelChain Lab's flagship AI-powered digital identity platform, BINGO, integrates these cutting-edge voice-processing AI models to solve real-world problems.
BINGO's voice-based verification system can be applied in several key areas, the company states:
- Privacy-enhancing Age Checks: With tightening regulations around age verification on social media and adult platforms, conventional protocols like entering one's date of birth are becoming inadequate. BINGO offers a voice-based age verification AI, which enables online platforms to verify different age groups with over 88% accuracy using just a 4-second voice clip.
- Instant Age Identification in Emergency Calls: In emergency call centres, every second counts. BINGO's voice-based age verification AI can instantly identify a caller's age group without the need to ask questions, enabling timely support.
- Fraud Prevention in Voice-based Banking Services: In voice-activated and telephone banking services, BINGO's voice matching and anti-spoofing capabilities reduce the risk of identity theft and bank account takeovers.
- Enhanced Security in Smart Homes and IoT: As smart home devices become more widespread, ensuring the security of voice-activated systems is critical. BINGO's voice verification AI can integrate into smart home ecosystems to prevent unauthorised users from controlling devices like smart locks, security cameras, and thermostats, thus enhancing overall security.
