Breakthrough in Using Call Data for Developing AI to Prevent Voice Phishing

6 Organizations Collaborate to Develop Phishing Prevention AI
Providing Real Call Data... Protecting Personal Information Together
First Corporate Case SKT "Utilizing On-Device AI"

Six government ministries and agencies will collaborate to develop artificial intelligence (AI) and services to prevent damage from voice phishing, a representative financial crime affecting the public. As a result, private companies will be able to utilize actual voice phishing call data for related service development.

On the 3rd, the Ministry of Science and ICT, Financial Services Commission, Personal Information Protection Commission, Financial Supervisory Service, National Forensic Service, and Korea Internet & Security Agency (KISA) announced that they signed a "Mutual Memorandum of Understanding for AI and Data-Based Voice Phishing Prevention" at the Government Complex Seoul.

Through this agreement, private companies such as telecommunications firms will be able to receive relevant call data when developing AI technologies and services for voice phishing prevention.

The Financial Supervisory Service provides voice call data collected from victim reports to the National Forensic Service, which then processes the data through de-identification and other preprocessing steps before providing it to the private sector. The Personal Information Protection Commission and KISA review legal interpretations and regulatory improvement measures regarding issues under the Personal Information Protection Act that may arise during data provision, collection, and use, and support comprehensive consulting on the use of pseudonymized information.

The Ministry of Science and ICT, Financial Services Commission, and Personal Information Protection Commission actively discover AI technologies and services for voice phishing prevention in cooperation with the telecommunications and financial industries. The Personal Information Protection Commission establishes compliance measures for personal information laws during service development and utilizes its "Preliminary Appropriateness Review System," which exempts businesses from disadvantageous dispositions under the Personal Information Protection Act if they comply.

Additionally, the Ministry of Science and ICT and the Personal Information Protection Commission plan and promote government-led R&D projects for voice phishing response technology development. The Ministry of Science and ICT plans specific projects with relevant experts and supports budgets. The Personal Information Protection Commission identifies regulatory improvements related to the Personal Information Protection Act during the research process and promotes regulatory sandbox measures if necessary.

Based on this inter-agency collaboration, SK Telecom will develop an AI service for detecting and preventing voice phishing. This service will include a function that determines in real time whether a call is suspected of voice phishing based on call context and notifies the user or their family.

For example, if key keywords or patterns commonly used in voice phishing are detected, or if the call context analysis identifies impersonation of investigative agencies or requests for personal information under the pretext of financial transactions, the call will be classified as suspicious. By analyzing call content with AI to detect voice phishing, it is expected to respond more effectively to voice phishing crime methods than simply blocking suspicious lines.

Conceptual Diagram of SK Telecom's On-Device AI-Based Real-Time Voice Phishing Detection

SKT plans to enhance personal information protection by using on-device AI technology that processes call data within the device without transmitting it to servers. This requires a small language model (sLM) that needs high-quality data. Accordingly, the National Forensic Service has prepared a plan to convert about 21,000 call data records into text while protecting the rights of data subjects, and to provide SKT with data after de-identifying sensitive information based on consultations with the Personal Information Protection Commission and KISA. Currently, pseudonymization of the data is underway, and the data is expected to be provided in June.

After receiving the data, SKT will refine the model’s performance through fine-tuning, verify it in prototypes, and then launch the service. They will continue to receive data to update the model in the future.

In addition, the Personal Information Protection Commission has received review requests related to voice phishing service development from multiple companies including telecommunications firms and plans to discuss further with the Financial Services Commission, Ministry of Science and ICT, and others.

The Ministry of Science and ICT is planning and promoting an R&D project for early detection technology of new types of voice phishing in 2025 and is discussing safe data utilization measures together with the Personal Information Protection Commission.

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