// security researcher · hong kong
Security researcher
“Every vulnerability discovered is a step towards a safer internet for everyone.”
Six production-grade tools spanning AI red-teaming, offensive-security benchmarking, and hardened systems software.
Autonomous Red & Blue Teaming — Attack and Defend with AI
An agentic red-teaming framework built on LangGraph and Re-Act. It chains recon, exploitation, and post-exploitation into one pipeline, while a Blue Team SOC defends live traffic with 18 attack-pattern detectors and AI-driven countermeasures.
Execution. Not Theory.
A production-grade benchmark that measures whether an LLM can actually execute the full kill chain against live targets — not recite CVEs. Four stages, five vulnerable labs, real tools, real flags.
Never accidentally destroy a disk again.
A memory-safe reimplementation of Unix dd with a 5-layer safety system, zero-copy kernel transfers, and both legacy key=value syntax plus modern subcommands. Fast, auditable, and safe by default.
AES-256 at every layer. Zero trace when locked.
A portable encrypted file store for macOS and Linux with dual-layer AES-256 encryption, multi-vault support, and a disguise mode. When locked, nothing remains — no plaintext, no metadata, no trace.
Write a human-readable sentence and the cron will still work.
A drop-in replacement for Vixie Cron that understands plain English. Over 250 natural-language patterns map to classic cron expressions while preserving every operator, special string, and edge case.
A lifesaving tool for people locked out of Macs they own.
Documents a physical-access attack chain on macOS — and the mitigations that stop it. A LaunchDaemon masquerading as an Apple service persists root access across reboots, with full detection guidance for blue teams.
Every layer chosen for a reason — memory safety, observability, portability, and speed.
Rustdd-rs
PythonMedusa · BreachBench
Ccrontastic · dd-rs
ShellMacVault · evilmaid
LangGraphMedusa
DockerBreachBench labs
MITRE ATT&CK80+ mappings
AES-256MacVault
APFS / LUKSMacVault
Rich TUIMedusa
MermaidAttack chains
LLM providersDeepSeek · Gemini · Anthropic“Every vulnerability discovered is a step towards a safer internet for everyone.”
I don't break things for their own sake. I find the cracks before they are used against real people — then I build tools that make powerful technology safe, accessible, and useful. Responsible disclosure and ethical research power everything I do.
I ship benchmarks that run live exploits, not multiple-choice quizzes. If I can't demonstrate it against a real target, it doesn't count.
My findings are documented, reproducible, and shared with defenders. Every attack path ships alongside the mitigations that close it.
From a dd that can't nuke your disk to a vault that leaves zero trace, I trade cleverness for real-world safety.