Financial Security Institute to Use AI to Assess Voice-Phishing Risk and Share Results with Financial Companies

AI Detection Model Integrated into ASAP
Fraud Detection Capabilities of Three Internet-Only Banks Combined
Small and Midsize Financial Companies Can Use It Without Building Separate Systems

The Financial Security Institute is launching a service that uses artificial intelligence (AI) to assess the risk of suspected voice phishing transactions and provide the results to financial companies.

The Financial Security Institute announced on November 9 that it has integrated an AI-based anomalous financial transaction detection model into ASAP, its AI platform for sharing voice phishing information, and is offering a service that allows financial companies to check the "AI Phishing Index."

The AI Phishing Index is a metric that uses AI to assess and score the likelihood that a financial transaction is an actual voice phishing case. Financial companies can refer to the score to assess the risk of suspicious transactions and use it in their efforts to combat voice phishing. They can also use a large language model (LLM) to further analyze the AI Phishing Index and the key factors used to calculate it, including estimating why each factor affected the risk score, for additional analysis.

The service uses an AI model that incorporates the fraud detection capabilities of three internet-only banks: KakaoBank, Toss Bank, and K Bank. It applies federated learning, a technology that enables the creation of AI models that can be used collaboratively while maintaining privacy and security, without directly sharing each bank's data.

The Financial Security Institute will operate the AI-based anomalous financial transaction detection model and its infrastructure directly. Financial companies can access the service through an application programming interface (API), subject to set usage limits. The technology is also expected to be readily accessible to small and midsize financial companies, without the burden of building their own infrastructure or operating AI models.

The service is being developed as part of efforts to upgrade voice phishing detection and response systems under the measures to strengthen security management across the financial sector, announced by financial authorities on November 5.

The Financial Security Institute will post a video explaining how to use the service for institutions participating in ASAP through November 13. After monitoring the service, it plans to determine appropriate usage limits and identify areas for improvement, and actively support financial companies in using the service.

"We developed this service so financial companies can easily use AI models in their voice phishing response work without the burden of building infrastructure or operating AI models," said Park Sangwon, head of the Financial Security Institute. "We will help strengthen AI-based voice phishing response capabilities across the financial sector, including among small and midsize financial companies."

This content was produced with the assistance of AI translation services.