Summary
High-school security researcher and generative-AI developer focused on AI/LLM security, web security, penetration testing, and red-team automation. Independently completed a prompt-injection detection study (30,000+ trials, formally published) and built a multi-agent automated offense/defense system that placed in the upper tier of domestic and international CTFs. Inventor on three R.O.C. utility-model patents; trainee of the Ministry of Education AIS3 program; served as red-team operator and security assessor in national cyber offense/defense drills.
Skills
- LanguagesPython, C++
- SecurityWeb security, cryptography, penetration testing, AI/LLM security, CTF
- AI / DevelopmentGenerative-AI & LLM applications, multi-agent systems, web front-end, hardware prototyping
- OtherNetwork administration, project management, cross-disciplinary integration, technical writing & teaching
Experience
Red-Team Operator / Security Assessor — National Institute of Cyber Security & Ministry of Education2025
- Red-team operator in the Executive Yuan national cyber offense/defense drill; assessor in the education-sector security drill
CTF Challenge Author — FHFCTF, THJCC CTF2024 · 2026
- Designed web and cryptography challenges for two security competitions, covering scenario design, deployment, and solve validation
Part-time Staff — Emerging-Technology Distance Education Demonstration Program—
Projects
Multi-Model Collaborative Detection of Prompt-Injection Attacks (research paper)2026
- Proposed an architecture where three heterogeneous commercial LLMs vote to jointly detect attacks, covering the blind spots of any single model
- Reached ~87% detection on long text and ~99.8% on short instructions, outperforming every individual model, over 30,000+ trials
- Solely handled literature review, experiment design, implementation, and analysis; awarded Distinction in the National Senior-High Research-Paper Competition (Information) and presented at an academic symposium
Automated Red-Team Multi-Agent System2025–2026
- Orchestrator agent dispatches parallel worker agents by challenge category, with shared memory and self-evolving rules
- Three-tier decision model (hard safety floor / adjustable rules / model reasoning) plus a built-in authorization & ethics policy
- AIS3 Pre-Exam: 14th individually (3,912 pts); UMDCTF: 72nd overall, 22nd in the student division
Industrial Control System (ICS) Security Range — MOE AIS32025
- Built a low-cost physical ICS range on self-made hardware; analyzed OT protocols and designed a progressive attack/defense workflow
Trashform — Image-Recognition Recycling System (PWA)—
- Photograph an item to identify it and get disassembly & sorting guidance per each city's rules; built with the RETHINK NGO
Aircraft Marshaller — Ground-Marshalling Simulator—
- Real-time in-browser gesture recognition with 3D guidance, courseware for standard aviation ground-marshalling signals (no image upload)
Education
- Taipei Municipal Digital Experimental High School (T-School)2023–2026 (expected)
- Nanshan High School — Affiliated Junior High, New Taipei City
Awards & Honors
- 2025 IEYI World Youth Invention Exhibition, Taiwan Selection — Gold (Team Taiwan) / World Final — Silver
- National Senior-High Research-Paper Competition, Information — Distinction (2026)
- 2024 Golden Shield Cybersecurity Skills Award, Finals — Emerging Talent Award
- 2025 CDX Cyber Offense/Defense Exchange — 5th place
- UMDCTF (international) — Open 72/763, Student 22/146
- 2025 MyFirstCTF — 26th; 2024 CGGC Cyber Guardian Challenge — 27th
Certifications & Training
- iPAS Information Security Engineer (MOEA) — Intermediate
- MOE Advanced Cybersecurity Talent Program — AIS3, AIS3 Junior
- TAIWAN HolyHigh Training, 10th cohort — Advanced
- HackTheBox — 7+ Labs completed
Talks
- Security course instructor (school & external): AI/LLM offense & defense, cryptography, OSINT, Linux/Kali, web security / CTF fundamentals
- MediaTek STEM educators' AI workshop — Teaching Assistant
- COSCUP 2026 ("Hackers In Taiwan") — Speaker
Publications
- W.-C. Zeng, "Multi-Model Collaborative Detection of Prompt-Injection Attacks," 2026 Symposium on AI & Modern Management Theory and Practice — presented.
Patents (Inventor)
- R.O.C. Utility Model M669931 — Generative-AI-based intelligent web application firewall system
- R.O.C. Utility Model M650895 — Intelligent fossil identification system
- R.O.C. Utility Model M678878 — Smart elevator operation system
Leadership & Community
- T-School InfoSec Club — President; Aviation Club — Vice President
- ISIP inter-school security club "ZeroDay C00kie" — Course Lead; 2026 "Bak3 C00kie" camp — Chief Organizer
- NTUST Information Security Club — Member
- 2025 HITCON Cyber Range — Operations; 2026 SITCON — Operations