Sumsub, an all-encompassing verification platform, has today unveiled ‘For Fake’s Sake’, a collection of models designed to identify deepfakes and synthetic fraud in visual content. This move is part of Sumsub’s ongoing commitment to improve AI security and transparency.
The company has a rich history in scrutinising visual data, distinguishing genuine individuals from AI-generated imitations. Sumsub’s in-house AI/ML Research Lab has now broadened this expertise, developing four unique machine learning models to tackle deepfake and synthetic fraud detection.
Key Highlights of ‘For Fake’s Sake’:
- An innovative technical approach that empowers users to interact responsibly with AI-produced content.
- The ability to evolve alongside other AI-powered tools.
- Provides users with the capability to assess the likelihood of an image being artificially created.
Following AI community standards, Sumsub offers comprehensive information on trained models, including datasets and metrics. Although internal testing at Sumsub suggests that the models are highly effective in detecting standard image modifications, they can be even more potent when used in conjunction with other content analysis methods. This multi-pronged approach ensures the confident recognition of AI-generated images.
Vyacheslav Zholudev, co-founder and CTO of Sumsub, shared his thoughts on the initiative:
“We’ve invested significant efforts to counteract the threat of deepfakes and the harmful effects of synthetic fraud by continuously enhancing our existing solutions and incorporating new features, such as the improved deepfake detector of our liveness tool.”
He emphasised the importance of not only delivering a top-tier product to customers but also engaging in open discussions about the complex threats posed by deepfakes. These threats range from online impersonation and personal data theft to misinformation dissemination and social engineering.
Zholudev added:
“Our open-source tool is just one step in Sumsub’s journey towards our mission of fostering a secure and inclusive digital future—for businesses and the broader public alike.”
Sumsub will utilise feedback from the AI research community to hone the models’ capabilities further, demonstrating their commitment to constant improvement. This announcement signals Sumsub’s determination to lead the fight against deepfakes and synthetic fraud, bolstering security in an increasingly digital world.