Welcome!!!😘😘😘
I’m Yijie Lu(陆一杰)🐕. I’m an undergraduate student at the School of Cyber Science and Engineering, Wuhan University (WHU), majoring in Cyberspace Security.
My current academic standing is Rank: 5/143 (Top 3.4%) with a GPA of 3.82/4.0.
Over the past years, I have been honored with the First Class Scholarship of WHU (Top 5%) and the Lei Jun Computer Science Undergraduate Scholarship (60 recipients university-wide).
I look forward to connecting and communicating with more people, if you are interested in me, please drop me an email!☀️
🔥 News
- Aug 2024: Won Bronze Medal in the NeurIPS-Ariel Data Challenge 2024.
- Aug 2024: Won Bronze Medal in the Kaggle: LMSYS-Chatbot Arena Human Preference Predictions.
- Sep 2024: As Team Leader, won the First Prize (National Final) in the China Collegiate Computing Contest - Network Challenge.
- Jun 2024: As a core member, won the First Prize (National Final) in the China Robot and Artificial Intelligence Competition.
📝 Publications & Research
$^\dagger$: Equal contribution, $^*$: Corresponding author
- EVA: Red-Teaming GUI Agents via Evolving Indirect Prompt Injection
Preprint, submitted to NeurIPS 2025.
[arXiv]
This project proposes a black-box, feedback-driven evolutionary framework for red-teaming GUI agents against indirect prompt injection. By dynamically evolving a word corpus based on the agent’s interactive feedback, EVA significantly improves attack success rates (e.g., from 48% to 80% in popup scenarios) and reveals key vulnerabilities related to attentional focus and semantic overlap, offering actionable insights for defense.
- PoisonBench: A Fine-Grained Direct Injection Benchmark for Robustness Evaluation in Retrieval-Augmented Generation with Large Language Models
Preprint, submitted to AAAI 2026.
We introduce PoisonBench, the first bilingual (Chinese-English) benchmark focused on templated direct injection attacks in RAG systems. It features six fine-grained poisoning templates based on cognitive biases (e.g., authority citation, statistical misdirection). Our experiments on mainstream LLMs show a significant drop in accuracy (up to 27.4 percentage points), demonstrating that models’ internal knowledge is easily overridden by rhetorically strong but factually weak contexts.
- Smart Sound Guidance - IoT Voice Attack Defense System
[Project Homepage] | [Code]
This project addresses the deficiencies of traditional voice attack defense systems. We designed a comprehensive solution featuring an intelligent attack detection model (88.7% accuracy), a crowdsensing monitoring platform, and a threat warning system, effectively enhancing the security of IoT devices against voice-based threats. This work, which I led, won the First Prize in the National Final of the China Collegiate Computing Contest.
- HaMonitorSentry - High-rise Building Intelligent Monitoring System
[Project Homepage] | [Code] | [Docs]
Focusing on dual-angle monitoring, this system integrates algorithms for detecting falling objects and recognizing dangerous high-rise behaviors. We introduced techniques like inter-class distance knowledge distillation and a Mask-Attention module to improve detection accuracy and real-time performance in complex scenes. This work won the First Prize in the National Final of the China Robot and Artificial Intelligence Competition.
- FuzzLLM - A Novel Universal Fuzzing Framework for Actively Discovering LLM Jailbreak Vulnerabilities
[Project Homepage] | [Code]
Inspired by traditional fuzzing, FuzzLLM deconstructs jailbreak prompts and automatically generates a large volume of structured test cases. This highly automated framework can efficiently identify potential security vulnerabilities in LLMs in a black-box setting, enabling proactive defense. The project has resulted in one National Invention Patent and one Software Copyright.
🏆 Competition Awards
- Bronze Medal. NeurIPS-Ariel Data Challenge 2024. (International, 2024)
- Bronze Medal. Kaggle: LMSYS-Chatbot Arena Human Preference Predictions. (International, 2024)
- First Prize. China Collegiate Computing Contest - Network Challenge (National Final). (National, 2024)
- First Prize. China Robot and Artificial Intelligence Competition (National Final). (National, 2024)
- Second Prize. National University Business Elite Challenge - Innovation and Entrepreneurship Competition (National Final). (National, 2023)
- Second Prize. LanQiao Cup National Software and IT Professionals Competition - Digital Technology Innovation Contest (National Preliminaries). (National, 2024)
- Third Prize. LanQiao Cup National Software and IT Professionals Competition - AI Programming Contest (National Preliminaries). (National, 2025)
- First Prize. China Robot and Artificial Intelligence Competition (Hubei Provincial). (Provincial, 2025)
- Second Prize. China College Student Computer Design Competition (Central South Region). (Provincial, 2025)
- Second Prize. China Collegiate Computing Contest - Network Challenge (Central China Region). (Provincial, 2024)
- Second Prize. China College Student Computer Design Competition (Central South Region). (Provincial, 2024)
- Second Prize. China Robot and Artificial Intelligence Competition (Hubei Provincial). (Provincial, 2024)
- Third Prize. China Robot and Artificial Intelligence Competition (Hubei Provincial). (Provincial, 2025)
- Third Prize. China College Student Computer Design Competition (Central South Region). (Provincial, 2024)
- Third Prize. China College Student Computer Design Competition (Central South Region, AI Track). (Provincial, 2023)
- Third Prize. China College Student Computer Design Competition (Central South Region, Digital Media Track). (Provincial, 2023)
- Third Prize. RAICOM Robot Developer Competition (Hubei Provincial). (Provincial, 2024)
- Third Prize. Huazhong Cup Mathematical Contest in Modeling. (Provincial, 2024)
🎖 Scholarships and Honors
- 2023.10 Lei Jun Computer Science Undergraduate Scholarship (¥10000, 60 recipients university-wide) Wuhan University & Xiaomi Inc.
- 2023.10 First Class Scholarship of WHU (Top 5%) Wuhan University
- 2023.10 Merit Student (Top 10%) Wuhan University
- LvMeng Scholarship Wuhan University
- Advanced Individual in Scientific and Technological Innovation Wuhan University
📖 Educations
- 2022.08 - Present, B.E. in Cyberspace Security, Wuhan University, China.
-
GPA: 3.82 / 4.0 Rank: 5 / 143 (Top 3.4%) - Core Courses: Advanced Mathematics (91), Discrete Mathematics (91), Artificial Intelligence (96, 98), Machine Learning (90, 92), Security Maker Practice (98)
- Skills: C/C++, Python, HTML/CSS/JavaScript, Algorithms & Data Structures
- Tools: VS Code, PyCharm, Jupyter Notebook, LaTeX (Overleaf), Git
-
🎡 Activities & Services
- Teaching Assistant, WHU-Jisuanke Joint Practical Training Course “Security Maker Practice”, Jun 2025 - Jul 2025
- Awarded “Top TA” for contributions to teaching, exercise explanation, and Q&A sessions.
- Teaching Assistant, WHU-Jisuanke Joint Practical Training Course “Security Maker Practice”, Jul 2024 - Aug 2024
- Awarded “Excellent TA” for contributions to the one-month practical training course.
- Committee Member, Study Committee of the Class 3, Grade 2022, Aug 2022 - Present
- Member, Front-end Group, Ziqiang Student Network Culture Studio, Aug 2022 - Jun 2023