SHANGHAI: Deepfake technology, an artificial intelligence tool capable of generating convincingly fake audio and video, is increasingly being used to perpetrate financial crimes worldwide, raising serious concerns about sophisticated fraud.
In a notable incident reported by CNN earlier this year, a finance worker was tricked into transferring US$25 million during a video call with an individual posing as the company’s chief financial officer (CFO), who was actually a deepfake. Such an incident has intensified fears about the vulnerability of financial systems to advanced fraud techniques.
Furthermore, global fintech platforms are confronting a rising wave of AI-driven criminal activities. FinVolution, a leading fintech company, has reported an increase in AI-generated attacks on its platforms and has significantly invested in deepfake detection technologies to combat this threat.
The increasing prevalence of deepfake technology in financial crimes has been underscored by a report from Sumsub, an identity verification provider.
Its latest annual report revealed that identity fraud cases involving deepfakes have increased tenfold from 2022 to 2023. The situation in the Philippines is particularly concerning, with a staggering 4,500% increase in attempted fraud schemes utilising deepfake technology.
In China, identity fraud involving voice manipulation has outpaced facial deepfakes, with FinVolution intercepting over 1,000 such incidents in just a few months last year.
Meanwhile, Southeast Asia is experiencing a surge in AI visual deception techniques, such as facial swaps, which pose new challenges to the security of digital financial services.
FinVolution Vice President Lei Chen and head of its big data and AI division emphasised the urgency of the situation.
“Globally, the technology to detect fake voices is not keeping pace with the technology used to create them. We are pushing for advancements in AI that can detect these fakes, aiming to align these defences with the capabilities of large-scale model applications,” Chen said.
“Such efforts are vital for effectively safeguarding the security of public information and individual rights,” he added.
In an effort to combat these threats, FinVolution Group has heavily invested in developing voiceprint recognition anti-fraud solutions tailored for financial scenarios.
The company has taken a proactive approach by introducing its proprietary voiceprint recognition algorithmic model, which has been commercially utilised 2 years before external open-source models.
The model has gained recognition within a mere 4 seconds across millions of transactions. Moreover, it supports multiple languages, including Indonesian, Chinese, Spanish, and more, and holds a particularly strong position in Indonesian and Spanish markets.
FinVolution is also at the forefront of combating fraud in global financial markets with its tailor-made AI anti-fraud technologies. These cutting-edge services include advanced facial and document forgery detection and voice synthesis algorithms, which are integrated into apps of leading international brands.
By leveraging facial recognition and voice verification, these AI-driven tools play a crucial role in preventing illegal impersonation and bolstering the effectiveness of risk management strategies.
Notably, in Southeast Asian markets, FinVolution’s technologies stand out by accurately identifying and intercepting financial fraud activities with generative AI, achieving a detection accuracy rate of over 98%.
In another proactive move to advance AI deepfake detection development, FinVolution is leading the charge in fostering industry collaboration. This includes hosting competitions and supporting academic research.
For example, the company’s latest initiative – the 9th FinVolution Global Data Science Competition – zeroes in on deepfake speech detection and challenges global participants to leverage deep learning and AI adversarial techniques.
This competition targets the accurate identification of falsified speech generated by the latest large-scale models, with increasing difficulty levels reflecting evolving threats.
Notably, this year’s competition has been featured as part of the International Joint Conference on Artificial Intelligence (IJCAI) 2024 challenges.
Looking ahead, FinVolution remains steadfast in its commitment to advancing deepfake recognition technologies, prioritising user safety, and fostering a secure financial environment on a global scale.
