Recent reports highlight a startling surge in artificial intelligence-powered deepfake scams. In the first quarter of 2025 alone, documented losses exceeded USD 200 million USD, affecting both high-profile executives and private individuals. Deepfakes are increasingly leveraged for industrial-scale fraud through image, voice, and video impersonation.
In one notable 2023 case, scammers used AI-generated deepfake videos to impersonate a CFO at a Hong Kong company, tricking an employee into transferring USD 25 million to a fraudulent account.
FinVolution, a leading fintech company, has encountered similar threats firsthand. In 2023, the company detected a surge of deepfake-based face swap attacks in its Southeast market, where fraudsters attempted to bypass KYC (know-your-customer) systems of its lending platforms using manipulated video content. In response, FinVolution deployed a proprietary detection algorithm, achieving a 98% accuracy rate in identifying such fraudulent attempts.
Since then, the battle between generative AI and detection technology has only intensified. Deepfake images today exhibit increasingly subtle signs of manipulation, and artifacts once visible to the naked eye are now almost undetectable. However, algorithmic analysis continues to offer promising detection capabilities. In March, FinVolution upgraded multiple innovative algorithms capable of identifying forgery clues in both image backgrounds and facial regions.
Looking ahead, FinVolution’s experts anticipate that deepfake detection will move beyond static image analysis toward multimodal verification, such as detecting inconsistencies in both facial features and synthesized voice in video. The next generation of detection tools will also dive deeper into forensic-level detail, pinpointing manipulation traces and reverse-engineering the generative models used.
In addition to technical detection, the financial industry is also enhancing its KYC processes with dynamic, fraud-resistant methods such as randomized speech prompts, screen shimmer analysis, and behavioral app-state monitoring.
Contestants are tasked with implementing model training and inference on an image dataset provided by the organizer, which includes determining the probability of forgery in these images. The dataset, sourced from public repositories, consists of over 100,000 facial verification images designed for deepfake detection.
The competition has already drawn 426 teams and over 652 participants globally, including AI researchers and engineers from Microsoft, Xiaomi, Lenovo, Ant Group, the University of California, New York University, the University of Philippines, the University of Indonesia, and De La Salle University. Finalists will be announced in mid-September.
About FinVolution Group
FinVolution Group is a fintech company that connects millions of consumers as well as small and medium enterprises with financial institutions.
Founded in 2007 and listed on the New York Stock Exchange in 2017, we have been at the forefront of the pan-Asian credit technology industry, pioneering innovative technologies in credit risk assessment, fraud detection, big data, and AI. With a proven track record of robust growth in pan-Asian countries, we have established leading fintech platforms in China, Indonesia, the Philippines, and Pakistan.
This press release was published in partnership with FinVolution Group.