The announcement by cybersecurity firm McAfee at the 2024 Consumer Electronic Show (CES) of its Deepfake audio detection technology, known as Project Mockingbird, comes as no surprise to ValidSoft.
As a world-leading speech technology company, we have long predicted that synthetically generated audio, made possible by advancements in Artificial Intelligence (AI), would create a new form of virus that would be exploited by bad actors in various scenarios.
ValidSoft has been at the forefront of synthetic speech detection for years and developed its first such applications to strengthen its voice biometric systems in around 2012, well before the term Deepfake was even coined. In addition, ValidSoft participated in the multi-billion Euro, EU-funded, Horizon 2020 project since 2015, focusing on detecting synthetically generated speech. ValidSoft also collaborated with some of the world’s leading academic institutions on this cutting-edge R&D work, including working directly with the institutions behind the most famous series of academic challenges for fake audio detection, ASVspoof.
The Arms Race in AI: From Creation to Detection of Deepfakes
Our anti-spoof detection capabilities, particularly our Deepfake detection solutions, are based on years of experience and research in speech science, signal processing, and voice biometrics and deploy a combination of complex machine learning, AI, and large-scale Deep Neural Network (DNN) technologies.
In the same way that more traditional viruses and virus detection providers are locked in an arms race, we believe the same applies to AI generally and Deepfake detection specifically, meaning that although AI is the technology that develops deepfakes, it’s also the tool required to detect them. For ValidSoft, the key lies in our expertise in voice biometrics and signal processing to optimize the detection models. Our ongoing research and refinement of these large-scale DNN models is key to staying ahead in this arms race.
Redefining Fraud Prevention
Importantly, ValidSoft’s Deepfake detection tool Voice Verity™, is not restricted to analyzing videos for deepfake audio detection. ValidSoft has long predicted that the ability to generate synthetic audio would be exploited for fraudulent purposes in previously unseen use cases and has been custom-built to detect deep-fakes across a multitude of nefarious Use Cases that target both the enterprise and the consumer.
It is, for this reason, that Voice Verity™ integrates with any channel that supports audio, either real-time streamed or “at rest”. Targeted impersonation of people such as company officers and employees, for instance, over digital communication channels will undoubtedly provide bad actors with opportunities for fraud, and this has already been widely reported.
The virus malware landscape has changed forever and the AI deepfake war has begun. Given the inventiveness of fraudsters and the availability of generative AI tools, we believe that it is imperative that omni-channel deepfake detection will form an integral part of every organization’s anti-fraud defenses.
