Competition held to stop voice deepfakes scams
FinVolution held the Global Data Science Competition in Shanghai to tackle the growing threat of voice deepfakes and related online scams. Researchers, university representatives, and tech giants including Alibaba joined forces in this international effort.
The rapid advancement of deepfake technology blurs the lines between reality and AI-generated voices, posing a significant risk to personal privacy and financial security. Deepfakes can be used to create remarkably realistic voice impersonations, potentially tricking victims into revealing sensitive information or transferring money.
"Deepfake voice detection currently lags behind facial recognition accuracy," says Lv Qiang, an algorithm scientist at Shanghai-based FinVolution and a competition judge. "We're seeing around 70-80 percent accuracy for voice detection, compared with99 percent for facial recognition."
Imagine receiving a call that sounds like a friend or family member, only to discover it's a deepfake designed to steal your money. This scenario underscores the urgency of developing more sophisticated voice authentication methods.
Telecom fraud has been a long-term concern nationwide. There are even groups in Myanmar and Laos targeting Chinese people with phone scams.
Winning teams gather in Shanghai after showing their solutions to detect voice deepfakes.
Competition participants harnessed AI technologies, building models capable of detecting fake voices within provided data sets. The top team achieved a remarkable accuracy rate of more than 99 percent in the initial round, dropping to nearly 80 percent in the semifinals. Lv attributed this difference to the varying complexities present in the data sets.
"Some contestants used LLMs (large language model) to detect deepfakes generated by other LLMs, demonstrating promising results. These approaches have significant potential for the fintech sector," said Chen Lei, vice president of FinVolution and Head of Big Data and AI Division.
New York-listed FinVolution offers online financial services with annual trade volume of more than 200 billion yuan (US$27.8 billion) on its platform. Recognizing the dangers of deepfakes, the company has heavily invested in AI solutions to bolster financial security.
Chen and Lv were judges, along with others from Zhejiang University, Shanghai Jiao Tong University and Renmin University of China.
Launched in May, the 9th FinVolution Global Data Science Competition attracted 461 teams and 709 participants from leading universities and companies worldwide. The winning teams will attend the IJCAI (International Joint Conference on Artificial Intelligence) 2024, a top international AI conference held in Jeju, South Korea.
