San Francisco – June 3, 2025 – Paravision, a leader in trusted Identity AI, today announced the launch of Deepfake Detection 2.0, a major evolution in its identity fraud prevention suite. This new release significantly expands the company’s detection capabilities beyond traditional deepfakes, such as identity and expression swaps, to include synthetic faces generated by advanced AI models.

With this release, Paravision Deepfake Detection now delivers advanced accuracy in identifying not only traditional deepfakes, such as face swaps and expression swaps, with up to 97% improvements in error rates compared to version 1.0, but also entirely AI-generated faces, including those created using advanced diffusion and GAN technologies. These types of synthetic faces have become increasingly prevalent in fraud schemes, where attackers use them to impersonate nonexistent individuals or bypass onboarding and Know Your Customer (KYC) processes at scale.

We’ve built a foundation of trusted identity with industry-leading Face Recognition and Liveness Detection; technologies that are already securing critical systems around the world. With Deepfake Detection 2.0, we’re adding another essential building block, giving our partners the ability to detect not just who someone is, but whether the image itself is real. As AI-generated content becomes more sophisticated, our goal is to stay ahead of the threat curve and provide the most complete, secure, and future-ready identity infrastructure on the market.
Doug Aley, CEO of Paravision

Paravision Deepfake Detection works by analyzing facial imagery and generating a score that reflects the likelihood the image is authentic. The technology also provides scores for specific types of manipulation—differentiating traditional deepfakes from fully synthetic faces—giving partners tools to understand the prevalence and nature of deepfake attack vectors. Images scoring below a configured threshold are flagged as deepfakes, while those above are considered authentic, with thresholds customizable to fit specific use case requirements.

As attacks become increasingly sophisticated—both physical and digital—there’s no single silver bullet for identity fraud. With Face Recognition, Liveness, and Deepfake Detection, our identity toolkit powers a true defense-in-depth approach and helps our partners prepare and respond to the evolving threat landscape. And as threats continue to evolve, so will we, continuously advancing our technology to stay ahead.
Joey Pritikin, Chief Product Officer at Paravision

The new Deepfake Detection toolkit is available through Paravision’s cloud-ready SDKs and Docker-based architecture, making it easily deployable in digital onboarding, secure access, travel identity, and remote verification environments.