Scale of GenAI threat drives quest to develop the best deepfake detection tools

They can go anywhere and target anyone, from seniors to financial workers to students. They’re harder than ever to spot, and they can appear any time you interact with a screen or digital content feed. They’re deepfakes – the virtual zombies nibbling away at the border between truth and falsehood, threatening to unleash a digital chaos in which no one knows who’s real and who’s just a cluster of manipulated pixels.

The threat has drawn the attention of some of the world’s top experts in computing and forensics, who have launched firms aimed at providing the strongest possible defenses against fraud and digital fakery.

Polyguard real-time deepfake detection brings NASA expertise to the task

New York-based Polyguard has launched its real-time deepfake detection tool. A release says the product provides “proactive protection” against AI-generated deepfakes, multi-channel fraud and impersonation attacks.

Polyguard, which was founded in 2024, claims an industry-first in its real-time inbound and outbound number blocking. Its tool also offers tamper-evident caller ID spoofing protection, identity-verified encrypted calling channels and secure video conferencing on mobile and desktop, with integrations for call center software and meeting platforms such as Zoom. Its 3D facial recognition and hardware attestation blocks digital fakes and device tampering.

Joshua McKenty, CEO of Polyguard, is a former Chief Cloud Architect at NASA, who co-founded OpenStack and was Tech Lead for Netscape 8.0, among other things. From his perspective, the deepfake detection tools available on the market are largely ineffective, in that they operate defensively, after an attack has already begun.

“Everyone from bankers to high school students are facing an onslaught of deepfake-driven scams, from voice spoofing to video impersonation,” McKenty says. “Detection alone isn’t enough – by the time you realize a trusted caller ID is fake, the damage is done.”

Rather than “rely on AI battling AI,” which can inadvertently train generative adversarial networks (GANs) to create even more convincing deepfakes, Paravision says it proactively verifies identity to stop fraud before it starts.

Its deepfake detection is now available to financial institutions, family offices, professional advisors and individuals.

Boasting top forensics talent, GetReal Security nets $17.5M in Series A

GetReal Security, a New York based firm focused on deepfake detection for governments and enterprises, says it has closed $17.5 million in Series A funding. A release says the round is led by global early-stage cybersecurity and AI investment firm Forgepoint Capital, with participation from Ballistic Ventures, which also led GetReal’s $7 million seed round and incubated the company until it emerged from stealth in 2024. Evolution Equity and K2 Access Fund also participated, along with strategic investors In-Q-Tel (IQT), Cisco Investments and Capital One Ventures.

The investment will underpin ongoing R&D and product development in forensic analysis for the growing market in biometric authentication, identity verification and deepfake detection.

The company’s board of directors is also getting new members in Forgepoint Capital Managing Director Alberto Yépez and GetReal Chief Science Officer Dr. Hany Farid.

Farid, a respected academic with decades of experience, who has often offered forensic services to media and legal teams, is a particularly valuable asset for the company. A piece in TechCrunch calls Farid “a pioneer in techniques for identifying when digital images have been doctored” and says “Farid was understanding the risks of deepfakes before the term had even come into existence.”

Farid says he has dedicated his career to “advancing digital forensics techniques and advising on cases where manipulated content has dire real-world consequences,” and says GetReal is “a significant step forward in our mission to create best-in-class content verification and authentication capabilities to navigate an era where synthetic content is increasingly indistinguishable from reality.”

“Developments in foundation models and generative AI have evolved the attack surface well beyond traditional networks and infrastructures,” Yépez says. “This means that the verification and authentication of digital content – whether text, images, audio, or video – is a critical new frontier of enterprise risk management as organizations have become prime targets for manipulation and exploitation.”

Yépez says GetReal stands apart from competitors in their “combination of world-class digital forensics and deep cyber domain expertise” – a sentiment echoed by company CEO Matt Moyhahan, who tells TechCrunch, “No one’s peering into this the way that Hany does. But Hany can’t scale. So we basically took Hany and tried to create a ‘Hany service’ in the cloud.”

The firm’s unified platform provides infrastructure to accommodate all modalities (image, audio and video) for both files and real-time digital communication streams, and includes a web interface, an API and integrations to run media analysis as a service. It features an “Inspect” tool specifically aimed at safeguarding high-profile executives from identity theft and spoofing; a “Protect” tool to screen media; and “Respond,” which triggers analysis from human teams at GetReal.

Reality Defender one of 10 firms chosen for Singapore innovation program

New York deepfake fraud fighter Reality Defender is among the rescind cohort of firms chosen to receive support from the CyberBoost Catalyse programme, an initiative by the government-led CyberSG Talent, Innovation and Growth (TIG) Collaboration Centre.

An article in the Straits Times says Reality Defender has been operating in Singapore since 2021 and plans to use the CyberBoost Catalyse programme to broaden its customer base in the country, focusing on call centres. The programme offers training courses and workshops and assists companies in raising funds and applying for grants and industry support.

Like Paravision and GetReal, Reality Defender’s model is designed for constant deepfake monitoring, to detect AI-generated audio, video, image and text “on a large scale and in near real time.” The tool can be integrated into telephony systems used by call centres and as a plug-in on commercial video conferencing tools for enterprise.

Other countries represented in the CyberBoost Catalyse program’s second cohort include the UK, Croatia and Singapore itself. Netrust, which specialises in secure digital identities and electronic transactions for governments, banks and enterprises in Asia, has been operating since 1997 and is “one of the most established companies in this cohort.”

Article Topics

biometric liveness detection  |  deepfake detection  |  deepfakes  |  fraud prevention  |  generative AI  |  GetReal Security  |  Polyguard  |  Reality Defender