Ping An Digital Bank tests AI for deepfake fraud detection
Hong Kong-based Ping An Digital Bank has completed testing of AI tools aimed at detecting deepfake fraud within the Hong Kong Monetary Authority's generative AI sandbox. The project focused on real-time detection using facial recognition and an "AI vs. AI" approach.
Ping An Digital Bank, a virtual lender based in Hong Kong, announced on October 15th the completion of a project within the Hong Kong Monetary Authority's (HKMA) second generative AI sandbox cohort. The initiative focused on testing new tools designed to detect deepfake fraud.
The project involved developing an Anti-Fraud Strategy Platform, in collaboration with OneConnect Financial Technology. This platform utilizes facial recognition during identity verification processes to identify manipulated or synthetic selfies in real time. Ping An Digital Bank reported that sandbox tests achieved a detection accuracy exceeding 96 percent.
The bank also explored an "AI vs. AI" approach, where AI systems are employed to identify fraud generated by other AI systems. This strategy addresses the growing sophistication of digital scams.
This development comes amidst rising concerns over the use of deepfakes in financial crimes. A notable case in Hong Kong in 2024 saw an employee at a multinational company tricked into transferring US$25 million through a video call scam potentially involving deepfake technology.