"Stopping Financial Fraud"... Financial Security Institute Promotes Joint AI Model Development for Financial Sector

The Financial Security Institute will utilize AI (artificial intelligence) technology to support the security evaluation and verification of financial companies. It also plans to develop a joint AI model for detecting financial fraud transactions within the financial sector. The intention is to expand the response to financial fraud transactions from individual companies to a joint financial community system.

The Financial Security Institute announced this plan on the 20th, stating that it will actively support the safe and rapid adoption of AI technology in the financial sector.

As the use of generative AI in the financial sector increases, innovative changes are expected; however, concerns have also been raised regarding information leakage and misuse due to security vulnerabilities, as well as biased outcomes resulting from AI’s skewed learning. In its annual report last December, the U.S. Financial Stability Oversight Council (FSOC) pointed out risks such as cyberattacks occurring during the use of generative AI in finance and decision-making distortions caused by data bias.

In response, the Financial Security Institute will leverage its expertise to evaluate the security measures of financial companies designated as innovative financial services for generative AI use this year, supporting safe AI utilization even in network separation exception zones. As of February, 49 services from 34 financial companies have been submitted.

It will also support security verification by conducting simulated attacks on AI models used by financial companies to identify vulnerabilities. Simulated attacks refer to techniques that deceive AI with manipulated queries to induce incorrect answers or behaviors. Currently, eight financial companies have requested inspections for nine services.

The Institute is also promoting the development of a joint AI model for detecting financial fraud transactions in the financial sector using the new AI technology of federated learning. Federated learning is a distributed learning technology that allows multiple participants owning data to jointly develop an AI model; each participant trains an AI model individually and shares only the trained AI models without sharing data, which are then integrated. The Financial Security Institute stated, "By combining the strengths of fraud detection systems operated by each financial company through this technique, the overall financial fraud response capability across the financial sector can be enhanced."

Additionally, it plans to provide an environment where financial companies can easily use open-source AI models and support revisions to AI guidelines in the financial field.

Park Sang-won, President of the Financial Security Institute, said, "The successful adoption and utilization of AI technology will determine the future level of digital competitiveness in the financial sector," adding, "This year, we will actively invest and expand our organization to cultivate experts and strengthen technological capabilities, ensuring seamless support for AI innovation."