2024.06.28. The Fair Economy — SANDS Lab (CEO Kim Gi-hong), a company specializing in AI-based cyber threat intelligence (CTI), announced on the 28th that its consortium has won the 'Cybersecurity AI Dataset Construction and Utilization Enhancement Project' ordered by the Korea Internet & Security Agency (KISA) under the Ministry of Science and ICT.
According to SANDS Lab, as attackers increasingly exploit generative AI and ChatGPT to develop sophisticated new attack techniques, there is a growing need for incident response systems that can quickly identify and respond to fake content generated by AI, manipulated information, and subtly distorted data.
Against this backdrop, it is necessary to build an updated and advanced AI dataset that maintains consistency and compatibility with existing datasets while reflecting the latest threat trends involving detection evasion and the abuse of generative AI, as well as to develop an AI model capable of responding to actual intrusion incidents. The project aims to activate proactive adoption and use in the cybersecurity field through these datasets and models, while driving continuous development.
SANDS Lab, which excels in cyber threat intelligence, AI, and big data utilization, formed a consortium specialized in building AI-based active response systems with Logpresso (CEO Yang Bong-yeol), a company specializing in security operations (SecOps) platforms.
The project will focus on constructing and updating two types of AI datasets, including those for the latest malware and threat profiling. It will also involve preparing environments for data collection and processing, strengthening quality verification and demonstration at demand sites, supporting sharing and utilization, and identifying ways to expand use and disseminate results.
The company plans to realize automatic updates of attack group information and automation of the latest trend information using generative AI technology, in addition to a process optimized for building cybersecurity datasets and generating AI models. It aims to develop an on-premise small large language model (sLLM) specialized for cybersecurity to verify meaningful results and possibilities through direct application in real security environments, while also presenting plans for adopting and utilizing generative AI technology in the field.
Upon completion, the project will provide private entities with threat data that significantly impacts digital safety, such as generative AI, APT attack tools, and ransomware, laying the groundwork for proactive responses to the latest threats. It will also verify effectiveness for demand sites struggling to acquire AI datasets or needing to introduce intelligent incident response technologies using AI, playing a key role in strengthening cyber threat response capabilities through data.
By building, sharing, and promoting the use of a large-scale AI dataset specialized in national cybersecurity, and by leading the discovery of methodologies and best practices for applying AI in incident response, the project is expected to establish a growth foundation for domestic AI source technology in cybersecurity and contribute to securing global competitiveness for Korean security technologies.
Kim Gi-hong, SANDS Lab's CEO and overall project manager, said, "This project requires building and utilizing an information generation automation system that can assess the explanatory power of attack situations and determine response priorities, moving beyond traditional methods of simply collecting and analyzing threat information. Going a step further, we will fuse next-generation generative AI technologies, including LLMs, with cybersecurity technologies to promote the utilization of AI datasets and maximize their value through effectiveness and quality verification."
The total budget for the project is 4.5 billion won (about $3.2 million). Considering the consortium's composition ratio, SANDS Lab will handle 90% of the total project scale. In terms of contract value, this represents an increase of about 8% compared to last year. The project duration is six months, and if construction is completed within the year, the entire amount is expected to be reflected in this year's sales.
0 Comments
No comments yet.