Projects²⁰

ROS 2 robot car with web control, live LiDAR scans, and obstacle detection.

Claude-assisted PDF, TXT, and DOCX summarization with project-based summary management, JWT authentication, a React frontend, FastAPI backend, and Kubernetes deployment manifests.

C/C++ source-code vulnerability analysis agents combining NVD CVE mapping, LLM finding proposals, and deterministic evidence verification, with JSON/Markdown/PDF reporting.

QR phishing detection and QR-code generation service with a Flutter client, Random Forest URL classification, a Flask inference API, and a Spring Boot backend.

Role: Frontend · AI

Android malware detection using structural clustering and adaptive thresholds (CMES 2026).

Android malware detection using Leiden communities and mixture of experts (ACM SAC 2026).

Related TrustZone-A research prototypes: information-flow-based ML-KEM partitioning and secret-isolation interfaces, plus Kyber split-design documentation and an OP-TEE client/TA scaffold.

Integrated Zephyr OS and Arm CCA through a custom shim layer to build an automotive virtual ECU research platform.

Role: Research Intern · Seoul National University

Experiment code and results for semantic cluster-guided Android malware drift detection.

Method-level Android API sequence extraction with Python and an Electron GUI.

Developed frontend and backend components of an automotive digital forensics tool to visualize vehicle events and support digital-evidence analysis. The linked repository contains the React/TypeScript frontend prototype for case and evidence management.

Commissioned by: Korean National Police Agency and Ministry of Science and ICT

Conducted mobile-platform AI security experiments and contributed to peer-reviewed manuscript preparation within the laboratory’s concept-drift workstream as part of a funded intelligent-building program.

Commissioned by: Korea Energy Technology Evaluation and Planning (KETEP)

Contributed literature analysis, experimental work, and peer-reviewed writing on mobile-platform AI security in support of the laboratory’s data-driven vehicle-forensics research and publication pipeline.

Commissioned by: National Research Foundation of Korea and Ministry of Science and ICT

Source-code security review platform combining Semgrep static analysis, Claude-assisted vulnerability explanations, and PDF reporting through Flask services, with a React/TypeScript/Electron interface.

Role: Team lead · AI/security

AI-assisted pronunciation training and visual feedback prototype developed for the Dankook Startup Hackathon.

Python prototype combining Lamport one-time signatures and Merkle trees.

EfficientNetB0-based crop leaf disease classification.

Product review collection with sentiment and emotion keyword classification.

Tkinter and Pillow GUI for adding transparent padding to PNG images.

CLI for converting PNG and SVG images into multi-resolution ICO files.