Hello, I am Professor Yoon In-soo from the Department of Electrical and Electronic Engineering at KAIST, and I am researching computer security, specifically hacking. The AI Cyber Challenge is a competition organized by DARPA in the United States, aimed at developing innovative technologies to defend against cyber attacks and predict security vulnerabilities in advance through advanced AI techniques. Specifically, it is a competition where teams develop systems that fully automate vulnerability detection and patching using artificial intelligence. This initiative started with the goal of maximizing the role of AI in the field of security through global competition. Team Atlanta is formed primarily by members from Professor Kim Tae-soo's lab at the Georgia Institute of Technology. I completed my doctoral studies in Professor Kim's lab, and Professor Park Sang-dong from POSTECH worked as a research fellow in that lab. Currently, Professor Kim Tae-soo is serving as an executive at Samsung Research.

Because you are gaining experience, we used communication tools such as KAIST, POSTECH, Samsung Research, and Georgia Tech to communicate in real-time, sharing the progress of the project and addressing situations that required immediate collaboration. Regular meetings were held via Zoom, allowing us to discuss in-depth the issues or ideas faced by each team member. Additionally, we efficiently managed our work and systematically operated the project by utilizing various development tools and frameworks. 

To highlight the strengths of each institution and school, the Georgia Tech research lab and KAIST research lab are teams conducting cutting-edge research in security vulnerabilities and defense techniques, playing a central role in this field. Professor Park Sang-dong from POSTECH is an outstanding expert in the field of AI security and has conducted research on aspects related to AI security. Samsung...

Research has helped our team based on practical experience gained from the industry and excellent development capabilities. It seems that the synergy of various team members came together to achieve good results. The core task of the competition was to build a system that could automatically detect security vulnerabilities using AI and correct them in real-time. To achieve this, our team focused on actively integrating existing large language models, such as GPT-4, with current software analysis technologies. As a result, we recorded achievements among the esteemed global teams that participated and advanced to the finals with a prize of $2 million. Leading such a large-scale project with many students was a first-time experience for me as a professor, so we faced various difficulties during the collaboration process, especially compared to research conducted individually.

In a project where people participate simultaneously and perform their respective roles, it was crucial to establish a system that allows everyone to share and verify the work done. In this regard, learning and mastering collaborative technologies was quite a challenge for both me and the students. It was not just about focusing on one's own code or research results; there was a significant learning experience in aligning and coordinating the entire team's direction. The most memorable episode was when our developed system discovered a new vulnerability in the well-known software, Eslight 3. The competition featured problems based on various famous software, and Eslight 3 was one of them, being one of the most commonly used database systems. The fact that our tool identified a vulnerability that had not been discovered before in such an important software highlights the potential of AI technology in enhancing security in the future.

이런 대회에 참여하는 것은 분명히 값진 경험이지만, 동시에 많은 도전과제를 안겨줍니다. 우리는 이 과정을 통해 기술적 역량을 한층 더 발전시키고, AI 보안 문제 해결에 기여할 수 있는 새로운 가능성을 탐색할 것입니다. 앞으로도 계속해서 혁신적인 아이디어를 바탕으로, 더욱 강력한 시스템을 구축하기 위해 노력할 것입니다.

It is a challenging endeavor. Nevertheless, I would like to express my deep gratitude to the students who voluntarily joined our team and worked hard. Over the next year, I sincerely hope that we can work diligently together so that this competition becomes a shining memory for the students and concludes with great results.

#이벤트 KAIST와 연합으로 뭉쳤다✨AI 사이버챌린지 TOP 7💻 I KAIST 윤인수 교수

한국과학기술원(KAIST)의 공식 유튜브 채널입니다.
This is the official YouTube channel of KAIST.