Kobe University Achieves Breakthrough in Fraudulent Fund Transfer Detection with Privacy-Preserving Federated Learning
Kobe, Japan – July 22, 2025 – Kobe University announced today the successful completion of a real-world demonstration experiment utilizing advanced privacy-preserving federated learning technologies, “DeepProtect” and “eFL-Boost.” This pioneering research has yielded a significant improvement in the detection rates of fraudulent fund transfers, marking a crucial step forward in safeguarding financial transactions.
The experiment, conducted by researchers at Kobe University, focused on addressing the critical challenge of identifying and preventing illicit financial activities, particularly fraudulent fund transfers. Traditional methods often rely on centralized data analysis, which can pose significant privacy concerns for individuals and institutions. Federated learning, a novel machine learning approach, offers a solution by enabling the training of machine learning models across decentralized data sources without the need to share raw data.
The newly developed technologies, “DeepProtect” and “eFL-Boost,” represent a sophisticated evolution of federated learning, designed to enhance both privacy protection and the accuracy of the resulting models. “DeepProtect” is engineered to provide robust privacy guarantees by incorporating advanced differential privacy techniques, ensuring that individual data remains confidential even during the collaborative training process. “eFL-Boost,” on the other hand, builds upon existing federated learning frameworks to optimize the efficiency and performance of model training, leading to more accurate and reliable detection capabilities.
The demonstration experiment involved the application of these technologies to a simulated environment mimicking real-world financial transaction data. The results were highly encouraging, with the integrated system of “DeepProtect” and “eFL-Boost” demonstrating a marked improvement in the recall rate for detecting fraudulent transfers. The recall rate, a key metric in classification tasks, measures the proportion of actual positive cases (fraudulent transfers) that were correctly identified. By increasing this rate, the technologies significantly reduce the likelihood of fraudulent activities going unnoticed.
This successful experiment underscores the immense potential of privacy-preserving federated learning in addressing complex security challenges within the financial sector. The ability to train highly accurate fraud detection models without compromising sensitive user data is a significant advancement. It paves the way for more secure and trustworthy financial systems, offering enhanced protection against the ever-evolving landscape of financial crime.
Kobe University’s commitment to cutting-edge research in artificial intelligence and cybersecurity continues to yield impactful results. This breakthrough in fraudulent fund transfer detection is expected to have far-reaching implications for financial institutions, regulators, and consumers alike, contributing to a safer and more secure digital economy. Further research and development are anticipated to explore the broader applications of these privacy-preserving federated learning techniques in other sensitive domains.
プライバシー保護連合学習技術「DeepProtect」「eFL-Boost」を活用した不正送金検知の実証実験を実施し、再現率向上を確認
AI has delivered the news.
The answer to the following question is obtained from Google Gemini.
Kobe University published ‘プライバシー保護連合学習技術「DeepProtect」「eFL-Boost」を活用した不正送金検知の実証実験を実施し、再現率向上を確認’ at 2025-07-22 05:00. Please write a detailed article about this news in a polite tone with relevant information. Please reply in English with the article only.