Big data analytics company S2W announced on September 14 that it will participate in the "Specialized AI Foundation Model Development (Cybersecurity Sector) Project" organized by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA).
S2W will join the project as part of the Naver Cloud Consortium. This consortium includes domestic companies such as Naver Cloud, LG CNS, and LG AI Research Institute, as well as key national infrastructure organizations like Korea Aerospace Industries (KAI) and Korea Hydro & Nuclear Power, along with major universities and research institutions such as Seoul National University and the Korea Advanced Institute of Science and Technology (KAIST).
The plan is to develop and validate a super-large security-focused model with a final structure of 700 billion parameters (700B) based on Naver Cloud's "HyperCLOVA X" and LG AI Research Institute's "ExaOne," optimized for the domestic environment, thereby enhancing the competitiveness of security AI in South Korea.
S2W will leverage its accumulated technological capabilities in threat intelligence (TI) and security-focused AI for this project. The company plans to participate in key development processes such as model training and validation, drawing on its expertise in building specialized cybersecurity language models and converting vast amounts of threat data into intelligence. Notably, S2W aims to enhance the model's completeness and applicability by incorporating hidden channel data that is difficult to obtain from the surface web.
As generative AI is increasingly utilized in both cyber offense and defense, the need for specialized AI models in the security sector is growing. The performance of Anthropic's cybersecurity-focused model, "Claude Mythos," which can autonomously identify and exploit vulnerabilities, has accelerated the advancement of AI-driven cyber threat and response technologies.
With the expanding role of AI in cybersecurity, the government has also initiated the development of its own AI models based on domestic technology and security data. This effort aims to reduce reliance on overseas general-purpose models and enhance the country's capabilities in responding to cyber threats.
Suh Sang-deok, CEO of S2W, stated, "We will contribute to strengthening security sovereignty and effectively responding to evolving cyber threats by bolstering our independent technological foundation."
The project will run for a total of 10 months, from this month until July of next year. The government will support 256 B200 GPUs, while Naver Cloud and LG will each contribute 4,000 B200 GPUs and 256 H200 GPUs, respectively. The consortium plans to develop two models with 700 billion parameters, including a defense model based on HyperCLOVA X and an attack model based on ExaOne, utilizing approximately 830TB of security and industrial data.
The developed models will be validated in seven key sectors, including power, finance, science and technology, telecommunications, semiconductors, defense, and aerospace, and will be released as open-source for commercial use.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.