M.S. Computer Science · University of Illinois Urbana-Champaign

DongJun Kim

Agentic security and cyber benchmarks for frontier models.

I build AI agents that find real vulnerabilities and benchmarks that measure the cyber capabilities of frontier models — from agentic vulnerability discovery in production Windows binaries (DeepBinary) to long-horizon security evaluation (SEC-bench Pro, CyberDual), now used by OpenAI, Google, DeepSeek, and NIST. This work grows out of Windows kernel vulnerability research at Enki WhiteHat (2024 — 2026).

Portrait of DongJun Kim
Champaign, Illinois

Research

Preprint · arXiv 2026

CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance

Chenyuan Yang*, Hwiwon Lee*, Jongseong Kim, Yihan Yang, DongJun Kim, Yinfang Chen, Lingming Zhang

Agents act as attackers that embed exploitable vulnerabilities in open-source code changes and as reviewers that try to catch and repair them; re-running the attacker's proof of concept decides whether the exploit survives. 1,088 valid attacks from six attackers against three agentic reviewers.

Preprint · arXiv 2026

SEC-bench Pro: Can Language Models Solve Long-Horizon Software Security Tasks?

Hwiwon Lee, Jiawei Liu, DongJun Kim, Ziqi Zhang, Chunqiu Steven Xia, Lingming Zhang

Long-horizon security tasks in million-line systems — V8, SpiderMonkey, and the Linux kernel — including vulnerabilities with over $1 million in cumulative bounties. Adopted for frontier-model evaluation by OpenAI (GPT-5.5-Cyber, GPT-6 Astra), Google (Gemini 4 Carbon), DeepSeek, and NIST CAISI.

* Equal contribution

Recognition

News

Experience & Education

Talks

CTF Challenges

Challenges authored for national competitions.

Selected CVEs

Highest-severity disclosures. 90 in total across Microsoft, Apple, and open source.

CVEComponentCVSS
CVE-2026-97814 Windows Text Shaping — Remote Code Execution 9.8 Critical
CVE-2026-70306 Microsoft Office SharePoint — Spoofing 9.3 Critical
CVE-2026-45504 Microsoft Exchange Server — Elevation of Privilege 8.8 High
CVE-2026-102225 Windows Text Shaping — Remote Code Execution 8.8 High
CVE-2026-70571 Microsoft Exchange Server — Remote Code Execution 8.1 High
CVE-2026-33840 Win32k — Elevation of Privilege 7.8 High
View all 90 CVEs →

Contact

Email
korea.smlijun@gmail.com
GitHub
@smlijun
LinkedIn
kdj-smlijun
Location
Champaign, IL, USA