Hi, I'm Jeonghyun Sim

I work on Agentic AI

M.S. student at HCRL, Korea University · Advised by Prof. Huy Kang Kim

Publications CV Email
01 biography

Jeonghyun Sim is an M.S. student in Information Security at the Hacking and Countermeasure Research Lab (HCRL), Korea University, advised by Prof. Huy Kang Kim. Jeonghyun’s research focuses on agentic AI, multi-agent systems, AI security, and trustworthy machine learning, with current work on making LLM-based multi-agent automation secure and reliable, from intent alignment in agent reasoning to defenses against jailbreaks and prompt injections.

Earlier work covers the adversarial robustness of deep learning: detecting adversarial examples with Grad-CAM, attention, and clustering, and building autoencoder-based defenses for image classification and object detection. Jeonghyun has also worked on security for autonomous and unmanned vehicles, including CAN intrusion detection datasets, adversarial attacks on AV perception, and RL-based attack/defense agents.

Jeonghyun received a B.S. in Industrial Security from Dankook University and conducted research at the Intelligent Security Technology Lab (ISTL) there.

02 research interests

Agentic AI

Building secure and reliable LLM agents that plan, reason, and act autonomously.

Multi-Agent Systems

Intent alignment and detecting intentional misalignment in the reasoning of LLM multi-agent systems.

AI Security

Auditing LLM vulnerabilities, including jailbreaks and prompt injections, and defensive alignment.

Trustworthy Machine Learning

Adversarial robustness of vision models, from attack detection to restoration for autonomous driving perception.

03 news all news →
  1. 🚗 Started the NRF project Threat Modeling and Cyber Attack Defense Simulation for Future Unmanned Vehicles, developing RL-based attack/defense agents.
  2. 📄 Technical report Autoencoder-based Denoising Defense against Adversarial Attacks on Object Detection is out on arXiv.
  3. 💡 Filed two Korean patent applications on adversarial example restoration for autonomous driving and cyber attack detection in UAV environments.
  4. 🎓 Started my M.S. in Information Security at Korea University (HCRL) under Prof. Huy Kang Kim, supported by the BK21 FOUR Graduate Research Fellowship.
  5. 🎉 Our paper on adversarial attack detection with Grad-CAM and clustering was published in Systems (MDPI).
04 selected publications all papers →
  1. Systems
    A Generalized Framework for Adversarial Attack Detection and Prevention Using Grad-CAM and Clustering Techniques
    Jeonghyun Sim and Hyun-Min Song
    Systems, 2025
    DOI Bib HTML
  2. arXiv
    Autoencoder-based Denoising Defense against Adversarial Attacks on Object Detection
    Min Geun Song, Gang Min Kim, Woonmin Kim, and 4 more authors
    2025
    Technical report, HCRL, Korea University
    arXiv Bib
  3. JKIISC
    Autoencoder-Based Defense Technique against One-Pixel Adversarial Attacks in Image Classification
    Jeonghyun Sim and Hyun-Min Song
    Journal of the Korea Institute of Information Security & Cryptology, 2023
    Bib
05 patents
  1. KR Patent
    High-Frequency Correlation-Based Transformer Model for Adversarial Example Restoration in Autonomous Driving Images
    Korean Patent Application No. 10-2025-0166315
    Filed Nov 6, 2025 · Pending
  2. KR Patent
    Multi-Channel Image-Based Cyber Attack Detection System Using Feature Group-Based Multi-Dataset Integration in UAV Environments
    Korean Patent Application No. 10-2025-0161718
    Filed Oct 31, 2025 · Pending

Let's connect.

Feel free to reach out — I'm always happy to chat about research and collaboration.