hemlix@icl:~$
$ whoami

HEMLIX

$ I teach machines to win, and build the rails they run on. 自动化一切竞技性任务。

// Yuxuan Sun — Computing Sophomore @ Imperial College London. UROP @ Imperial IHC Lab. ex-Security Researcher @ Tencent Xuanwu Lab. CTF player @ r3kapig. 孙宇煊 — 计算机系大二 @ 帝国理工学院。 UROP @ 帝国理工 IHC Lab。 前安全研究员 @ 腾讯玄武实验室。 CTF 队员 @ r3kapig。

0x01// currently现况
  • Researcher @ G.O.S.S.I.P, Shanghai Jiao Tong University — designed novel multi-target direct fuzzing. Building RedShift, an automated vulnerability discovery tool that discovers vulnerabilities, generates PoCs, and exploits them.
  • UROP @ Imperial IHC Lab — mentored by Mark Chen and Hongxiang Fan. Collaborating with Microsoft.
  • CTF player with r3kapig — 5th/955 @ LA CTF 2026. 8th/970 @ Srdnlen CTF. Building Auto-CTFer. Team ranked #1 on CTFtime 2025.
  • Building GamingMaster — a single shared-weight model that masters nearly all competitive games. Sub-project: AlphaGen, a game bot for generals.io using imitation learning → reinforcement learning pipeline. Achieved intermediate human-level play. Collab with a MIT student.
  • 研究员 @ G.O.S.S.I.P, 上海交通大学 — 设计了新型多目标定向模糊测试方法。 正在构建 RedShift,一个自动化漏洞发现工具,可自动发现漏洞、生成 PoC 并完成利用。
  • 本科生科研 (UROP) @ 帝国理工学院 IHC Lab — 导师 Mark Chen、Hongxiang Fan。与 Microsoft 合作。
  • CTF 队员 @ r3kapig — 5th/955 @ LA CTF,8th/970 @ Srdnlen CTF。 构建 Auto-CTFer。战队 CTFtime 2025 排名 #1。
  • 构建 GamingMaster — 一个共享权重的模型,旨在掌握几乎所有竞技游戏。 子项目:AlphaGen,基于 模仿学习 → 强化学习 的 generals.io 游戏 Bot。已达到人类中等水平。与一位 MIT 学生合作。
0x02// history历史
  • Security Researcher @ Tencent Xuanwu Lab — built CodeRetrX, a comprehensive agentic code auditing MCP toolkit featuring SOTA semantic code retrieval. Black Hat Asia 2026 Used it to power a vulnerability detection agent that found 20+ vulnerabilities, including CVE-2025-57801 (High, CVSS 8.6) which I verified.
  • RL-based fuzzing with contextual combinatorial multi-armed bandits @ Imperial College London Security and Machine Learning Lab.
  • Cryptographic accumulators research @ Tencent (2023-2024) — novel transparency system with Merkle Trees. Paper cited by Intel Labs.
  • Contributed to GmSSL, an open-source cryptographic library for Chinese national (ShangMi) algorithms.
  • B-Emulator & B-Arm Assembler — full ARM emulator with UART/timer/HVS + RL-based cache replacement. Imperial Best Overall Project Award 2025
  • Teaching assistant & mentored Chinese NOI Gold medalists on automated offensive security @ Tencent TSpark 2025.
  • 安全研究员 @ 腾讯玄武实验室 — 构建了 CodeRetrX,一个全面的智能体代码审计 MCP 工具集,具备最强的语义代码检索能力。 Black Hat Asia 2026 基于此驱动构建的漏洞检测 Agent,发现了 20+ 个高危漏洞,包括我验证的 CVE-2025-57801(High Severity,CVSS 8.6)。
  • 基于上下文组合多臂老虎机的强化学习模糊测试 @ 帝国理工学院 Security and Machine Learning Lab。
  • 密码学累加器研究 @ 腾讯 (2023-2024) — 基于 Merkle 树的新型透明性系统。论文被 Intel Labs 引用。
  • 为开源国密密码库 GmSSL 贡献代码。
  • B-Emulator & B-Arm Assembler — 完整的 ARM 模拟器(UART/定时器/HVS)+ 基于 RL 的缓存替换策略。 帝国理工 2025 最佳项目奖
  • 担任助教,指导中国 NOI 金牌选手学习自动化攻击性安全技术 @ 腾讯 2025 星火计划。
0x03// vulnerabilities漏洞
id finding描述 severity等级
CVE-2025-57801 Signature malleability in gnark EdDSA/ECDSA.gnark EdDSA/ECDSA 签名延展性漏洞。 CVSS 8.6
CVE-2026-32614 SM9 encryption forgery in gmsm.gmsm SM9 加密伪造漏洞。 CVSS 7.5
Credited Authentication bypass via credential disclosure in generals.io Version 31.0.0.generals.io Version 31.0.0 凭据泄露导致的身份认证绕过漏洞。 0-day
Alpen Labs Bug bounty awarded.获得漏洞赏金。 $1,200
Nexus Mutual Bug bounty awarded.获得漏洞赏金。 $500
0x04// research研究
AI for Security Vulnerability Discovery Fuzzing Static Analysis Competitive Game AI Deep Learning LLM Agents AI Infra
AI驱动的安全分析 漏洞挖掘 模糊测试 静态分析 竞技游戏 AI 深度学习 LLM 智能体 AI 基础设施
0x05// stats数据
USACO Platinum
#32
worldwide — 2023 US Open Contest全球 — 2023 US Open Contest
Codeforces
Master 2216
Kaggle
#95 / 2422
Poisonous Mushrooms
0x06// misc杂项
[human]
piano = "Level 10, CMA Grading Examination"
go = "5-Dan"
games = "Ultimate Champion in Clash Royale"
generals = "Top-20 in generals.io"
badminton = "enthusiast"
[human]
piano = "钢琴 · 中央音乐学院十级"
go = "围棋 · 业余五段"
games = "皇室战争终极冠军联赛"
generals = "generals.io 全球 Top-20"
badminton = "羽毛球爱好者"
0x07// contact联系
$ echo $EMAIL → click to copy点击复制