[Edaily Reporter Kim Hyun-ah ] East Security, a security subsidiary of ESTsoft Corp.(047560), is participating in a national-level “security AI” development project. The company will utilize malware and cyber threat data accumulated through the operation of the “Al-Yak” antivirus software and endpoint security solutions to train AI models, with the goal of developing proprietary AI models tailored to the domestic security environment.

East Security announced that the Naver Cloud Consortium has been selected as the final contractor in the cybersecurity sector for the “Development of Specialized Artificial Intelligence (AI) Foundation Models” project, which is being promoted by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA).

This project aims to secure an AI foundation model optimized for the domestic security environment to counter increasingly sophisticated cyberattacks. The goal is to reduce dependence on general-purpose AI models from overseas and secure AI capabilities tailored to domestic cyber threats.

East Security to Handle “Real-World Security Data”

East Security’s primary role is to provide data and analytical expertise accumulated from actual security operations.

The company will utilize the vast amount of malware data, threat intelligence (CTI), endpoint threat information, and breach analysis experience—all gathered through the operation of various security products and services, including Allyak—to train the AI model.

The performance of an AI model is not determined solely by the scale of its parameters. High-quality security data—such as how actual attacks occur and are detected, what types of malware are introduced, and how they were countered—is crucial.

East Security plans to contribute to the development of AI models tailored to the domestic cyber threat landscape by leveraging the data it has accumulated through its security products.

The Naver Cloud Consortium has secured approximately 830 TB of security data in total. This includes operational data from Naver Cloud and LG Corp., supplemented by data from national infrastructure providers and key industries such as KEPCO KDN, Korea Hydro & Nuclear Power, the Korea Financial Security Institute, KISTI, and LG Uplus.

East Security’s security data and analytical capabilities will also be integrated into this large-scale data ecosystem.

Preemptive Training with 4,000 GPUs… Development of Two 700B-Class Models

The Naver Cloud Consortium is accelerating development by deploying its own infrastructure ahead of government-provided GPUs.

Naver Cloud has proactively deployed 4,000 of its own B200 GPUs, separate from the 256 NVIDIA B200 GPUs provided by the government. With the addition of 256 LG Corp. H200 GPUs, the total number of GPUs available to the consortium is 4,256.

Based on this, the consortium is developing two security-specialized models with a 700B-class MoE (Mix of Experts) architecture, built on HyperClova X and ExaOne. The plan is to develop one model that enhances defensive capabilities and another that enhances offensive capabilities, thereby covering the entire lifecycle from cyberattack detection to analysis and response.

The developed models will undergo real-world validation across seven key sectors: power, finance, science and technology, telecommunications, semiconductors, defense, and aerospace. In particular, the consortium plans to secure technology that enables the stable operation of AI even within national critical infrastructure with restricted external network access and in closed-network environments.

Combining Experience in ‘AI Security’ Commercialization with a Domestically Developed NPU

East Security has already been working on integrating AI into its security products.

Recently, the company launched “AlYak Agent-based EDR,” which incorporates its proprietary AI foundation model, K-EXAONE, marking its entry into the commercialization of AI-based security services.

EDR (Endpoint Detection and Response) is a security solution that detects, analyzes, and responds to anomalous behavior occurring on endpoints such as PCs and servers. Integrating AI into this solution enables the automation of tasks such as analyzing numerous security events and determining response priorities.

If a domestically developed AI model specialized for cybersecurity is developed through this initiative, it will expand the foundation for integrating it with East Security’s existing security solutions.

The company is also pursuing integration with domestically developed AI semiconductors (NPUs). East Security has previously signed a Memorandum of Understanding (MOU) with FuriosaAI for the development of next-generation K-AI security solutions and technical cooperation. The strategy is to accelerate the establishment of a “Sovereign AI” infrastructure tailored to the domestic security environment by leveraging both security-specialized AI models and domestically developed AI accelerators.

Models

to Be Released as

Open Source… Expanding the Domestic Security Industry

The

models developed

through this project are scheduled to be released as

open source

.

The Naver Cloud Consortium plans to validate the models at national critical infrastructure sites and major industrial facilities, then expand the ecosystem so that domestic security companies can utilize them.

It will also pursue the commercialization of security AI services through the Naver Cloud Marketplace and provide support to small and medium-sized enterprises (SMEs) in detecting vulnerabilities and cyberattacks.

East Security will also contribute to the practical application and commercialization of AI models in security operations, leveraging its proprietary security data and field experience.

Jeong Sang-won, CEO of East Security, stated, “We are pleased to have been selected as a key partner in this project to build a cybersecurity-specialized AI foundation model.” He added, “We will mobilize all our capabilities—including our proven ability to commercialize AI security services, as demonstrated by the launch of A-YAK Agent-based EDR, as well as our unparalleled security data and analytical capabilities—to help elevate the nation’s cybersecurity capabilities to the next level.”