KT proposed guidelines and standards necessary for developing artificial intelligence (AI) models sp..

KT proposed guidelines and standards necessary for developing artificial intelligence (AI) models specialized in cybersecurity as international standards. It is about preparing common standards from data to be used for AI learning to things to consider in the process of model development and evaluation.

KT announced on the 21st that it proposed guidelines for developing cybersecurity-specific AI models as a new international standard work item at the International Telecommunications Union Telecommunications Standards Division (ITU-T) Information Protection Research Group (SG17) International Standards Conference held in Chongqing, China from the 7th to the 11th.

ITU-TSG17 is a research team that develops international standards such as cybersecurity, security architecture, personal information protection, and AI security with the participation of experts from governments, companies and research institutes from around the world.

KT's proposal this time is a guideline for developing AI specialized in cybersecurity using a general-purpose foundation model. We focused on establishing common criteria to be considered in the model development process so that AI models can be used to detect cyberattacks or analyze related information.

The proposal contains cybersecurity knowledge and datasets to be used in developing AI models, as well as technical considerations needed when developing security-specific models. Evaluation criteria throughout the model development life cycle and recommendations to continuously improve model performance after development were also included.

KT believes that if such international standards are established, companies and institutions can use them to develop and operate cybersecurity AI according to certain standards. It is explained that it can also be used to increase interoperability between different AI-based security technologies and improve threat detection, analysis, and response capabilities.

He also participated in establishing international standards for digital trust. ITU-TSG17 has developed a 'trust framework' technology report to define trust and reliability in an information and communication technology (ICT) environment and apply it to systems and services.

At this meeting, KT proposed amendments to supplement the key term definitions included in the report and to add components and requirements of the trust framework. KT explained that the amendment was supported by the UK and India, and that the issuance of the technical report was completed.

Kim Chang-oh, head of KT's information security office's personal information protection group, participated in the international standardization work. Managing Director Kim is the WP3 chairman in charge of securing reliability and security strategies and cooperation at ITU-TSG17. Yeom Heung-yeol, chairman of KT's personal information protection advisory committee, who chaired SG17 from 2016 to 2024, also participated in the standardization work.

In the future, KT plans to continue its international standardization activities in the fields of AI cybersecurity, personal information protection, and digital trust. We also discuss how to link trust frameworks and cybersecurity reference architectures in the future.

"In the AI era, cybersecurity, personal information protection, and trust should not be handled individually, but should be internalized together in AI, data, and services," said Kim, head of the group. "We will continue to contribute to global standardization of AI security, privacy, and trust."