The 9th FinVolution Global Data Science Competition presents LLMs in Voice Deepfake Detection

• Speech data augmentation and large language model recognition (LLM) impress the jury
• The final datasets contained the latest LLM-synthesized voices and newly recorded false voices, making detection more difficult.
• Competition highlights the increased risks of deepfakes due to new technologies

SHANGHAI, July 24, 2024 /PRNewswire/ — FinVolution Group, a leading fintech service provider, concluded the 9th FinVolution Global Data Science Competition on July 23As part of the IJCAI 2024 Challenge Track, a top international AI conference, this year’s event focused on using large language models (LLMs) to effectively detect deepfakes in voices.

The rapid advancement of deepfake technology is blurring the lines between AI-generated and real voices, raising concerns about privacy and financial security. Contestants used deep learning and adversarial AI techniques to develop models capable of detecting fake voices from the datasets provided. The finals featured a variety of algorithmic models and training approaches, including LLMs and traditional end-to-end detection technologies.

The winning team achieved an impressive detection rate of over 99% for fake votes in the preliminary round and almost 80% in the semifinals. Qiang Lyu, algorithm researcher at FinVolution, attributed the differences to the different complexity of the data sets.

“The preliminary round datasets mainly consisted of cloned voices generated by end-to-end text-to-speech (TTS) systems, which were easier to identify,” Lyu explained. “However, in the semifinals, the datasets included the latest LLM-synthesized voices, newly recorded fake voices, and even mixed real and fake voices in over five languages, including English, French, and Spanish. This complexity made detection more difficult.”

Using LLMs to respond to LLM-generated deepfakes

The judges were impressed by the participants’ innovative solutions and practical business implementations. Many used speech augmentation to expand their data sets and deployed the latest LLM recognition technologies to improve inference efficiency and reduce processing time.

Lei Chen, Vice President of FinVolution Group and Head of Big Data and AI, commented: “The competition presented a number of effective solutions for detecting voice deepfakes. Some participants even used LLMs to detect deepfakes generated by other LLMs, showing promising results. These approaches have significant potential for the fintech sector if further developed.”

Chen added: “Research on language LLMs must address academic challenges and then translate these solutions into practical applications to meet real-world business needs..“

Yuxiang WangCOO & CTO of FinVolution Group, concluded: “As deepfake generation technology evolves, detection methods must keep pace to achieve a balance between attack and defense. We hope that the FinVolution Global Data Science Competition will continue to push the boundaries of technology, nurture talent, academia and industry, and encourage new innovations.”

The IJCAI 2024 Challenge

The 9th FinVolution Global Data Science Competition, which started in May, attracted 461 teams and 709 participants from leading universities and companies worldwide.

The nine finalist teams presented their solutions during the on-site defense and demonstrated the diverse applications of deepfake detection technologies. Judges from top institutions, including China University, Shanghai Jiao Tong University and Renmin University of Chinaassessed the presentations.

As part of the IJCAI (International Joint Conference on Artificial Intelligence) Competitions and Challenges Track, a top international AI conference, the winners will attend the IJCAI 2024 in Jeju and present their solutions. South KoreaFinVolution is a proud sponsor of IJCAI 2024.

Since its launch, the FinVolution Global Data Science Competition has attracted nearly 10,000 researchers from leading universities and companies around the world. Previous topics have included “Dialect Distance Recognition in Speech,” “Product Recognition in Smart Display Cases,” and “Credit Services for Small and Micro Enterprises.”

About FinVolution Group

FinVolution Group is a leading fintech company connecting millions of consumers and micro and small businesses with financial institutions.

Founded in 2007 and listed on the New York Stock Exchange since 2017, we are pioneers of the pan-Asian credit technology industry and have developed innovative technologies in credit risk assessment, fraud detection, big data and artificial intelligence. With a proven track record of robust growth in pan-Asian countries, we have developed leading fintech platforms in China, IndonesiaAnd the Philippines.