B.S. in Industrial Security & Cybersecurity, Chung-Ang University (GPA 4.06/4.5, Expected Aug 2026)

Researcher @ Trustworthy AI Lab · Security Vision Intelligence Lab (SVIL)

KITRI BoB 11th — Grand Prix

Adversarial behavior of LLM agents, machine unlearning for security-critical systems, and infrastructure-as-code security. Interested in understanding how autonomous AI agents behave under adversarial conditions and building frameworks to make ML systems safely updatable.

Security Vision Intelligence Lab (SVIL), CAU (2025–) · PI: Prof. Seungmin Rho · Advisor: Prof. Sungwoo Park

Proposed sample-level exact forgetting as an alternative to tree-level replacement under concept drift. Empirically decomposed the contribution of forgetting mechanism (HOW) vs. selection strategy (WHAT) across 3,400+ runs on 4 NIDS datasets.

Outperformed Adaptive Random Forest in 10 of 12 drift scenarios (G-mean +8.6–31.4%, 10–19x speedup). Core algorithm implemented in Rust with PyO3 bindings.

suda

Trustworthy AI Lab, CAU (2026–) · PI: Prof. Hoki Kim · First author (ongoing)

A benchmark for measuring model- and agent-level bias in how LLM agents select attack strategies within adversarial web environments. Designed the isolated experiment environment, the automated evaluation pipeline, and a literature-grounded prompt suite; quantified strategy bias with statistical measures (Shannon entropy, HHI, Jensen–Shannon divergence).

CyBiasBench

CEO (2023–2025)

Built a Terraform static-analysis platform for IaC security. Scans for misconfigurations across AWS/NCP and generates ISMS-P compliance reports. Started as BoB 11th Grand Prix project, selected for government Cloud SW Development Program.

init-cloud · initcloud-scanner

Evolved into IaCDOG — Terraform visualization dashboard (D3.js + Go) with an IaC education platform.

a-wawa · aww-dashboard · eduiac

Discovered 4 vulnerabilities in commercial NAS devices, all assigned KVE IDs by KISA (KVE-2026-0116, -0271, -0272, -0275; max CVSS 9.8). Performed ARM binary analysis with IDA Pro; developed PoC exploits and completed responsible disclosure.

OSSCA (2024)

Fixed security vulnerabilities in Yorkie, a Go-based collaborative editing server. Configured Kubernetes deployment infrastructure.