로그인|회원가입|고객센터|기업교육 문의
페이지 맨 위로 이동
검색버튼 메뉴버튼

DEFCON 34

SK Shieldus EQST Ranks Fifth at DEFCON 34 AI CTF

Dong-A Ilbo | Updated 2026.08.10
Proving real-world problem-solving capabilities through AI agent–based autonomous vulnerability discovery and attack design
Advancing AI model vulnerability research and AI red teaming countermeasure technologies based on hands-on experience
Members of SK shieldus’ white-hat hacker group EQST participating in DEF CON 34 pose for a commemorative photo. Photo courtesy of SK shieldus
SK shieldus (CEO Min Ki-sik) announced that its white-hat hacker group EQST ranked fifth overall among more than 200 teams worldwide in the AI Village CTF at DEF CON 34, a global hacking and security conference. EQST was the only Korean team to place in the top five, demonstrating its AI security capabilities.

At this year’s DEF CON, held in Las Vegas, USA, the main topic was the changes that technologies capable of autonomous judgment and action, such as large language models (LLMs) and AI agents, will bring to the cybersecurity environment. Among a total of 38 villages, the CTF held in the AI Village required teams to build AI agents that conduct penetration testing based on a given AI model and use them to solve security challenges.

Unlike traditional CTFs where humans directly carry out attacks, this competition required participants to design AI agents that analyze problems, choose attack methods, and obtain flags. Teams competed by designing the structure and prompts of their agents to optimize attack strategies.

EQST optimized AI models by problem type and adjusted agent behavior, while analyzing difficulty and scoring to determine the order of attack. In particular, in the process of designing AI agents to autonomously identify and exploit vulnerabilities, the team drew comprehensively on its penetration testing experience and capabilities in security problem analysis.

Through this competition, EQST validated attack techniques using agentic AI in a real-world environment. The AI agent attack strategies and solution cases obtained during the event will be used for future research on AI service and model vulnerabilities, strengthening AI red team capabilities, and advancing security response technologies.

In addition, EQST formed a joint team with Naver Cloud, Rubyya Lab, and Finland’s H-T8 to compete in the DEF CON 34 CTF finals. The joint team, one of 12 finalists selected from 686 teams in the preliminaries, competed against security experts from various countries using a combination of Attack & Defense and King of the Hill (KOTH) formats.

EQST, a professional white-hat hacker group of about 200 members, is active in various areas including vulnerability research, penetration testing, and malware analysis. Recently, it has continued to achieve results in domestic and international competitions, winning the AI red team hacking contests “Judgment Day” and “DreamHack Hacking Defense Competition 2026,” placing second in “R3CTF 2026,” and third in “Midnight Sun CTF 2026.”

Kim Tae-hyung, Head of EQST Business Group at SK shieldus, said, “Placing in the AI CTF top five is a result that demonstrates the AI security technologies and practical problem-solving capabilities EQST has accumulated,” adding, “We will continue to expand research on AI vulnerabilities and red teaming to strengthen a proactive security response system.”

Choi Yong-seok

AI-translated with ChatGPT. Provided as is; original Korean text prevails.
Popular News

경영·경제 질문은 AI 비서에게,
무엇이든 물어보세요.

Click!