陆皓喆

陆皓喆

Peking University

Hello everyone, my name is Haozhe Lu (陆皓喆). I am currently a first-year Master’s student in Network and Information Security(网络与信息安全) at Peking University(北京大学), where I am a member of The Knowledge Computing Lab.

I received my Bachelor’s degree in Cyber Security from the College of Cryptology and Cyber Science(密码与网络空间安全学院), Nankai University(南开大学), with a minor in Finance. Feel free to reach out to me at 15058298819@163.com if you have any collaboration opportunities or shared interests.

My research interests lie in:

  • LLM Security: RAG Attacks and Defenses, Poisoning Attacks and so on.

  • LLM for Security: LLM for Vulnerability Mining, LLM for Penetration Testing and so on.

I have organized the Course Note，Lab Report for the Information Security major courses at Nankai University, as well as the Final Exam for theoretical courses. So far, these resources have assisted over 1,000 students and accumulated more than 200,000 views.

📝 Publications

arxiv / Preprint
ToxicRAG

ToxicRAG: Compromising Retrieval-Augmented Generation Systems via Single-Shot Knowledge Poisoning Attacks

first author 2026.04 – 2026.09

  • Reviewed over ten papers and reproduced poisoning attacks such as PoisonedRAG, AuthChain, and CorruptRAG to establish baselines for our proposed attack.
  • Designed ToxicRAG, a single-text "one-shot" attack that integrates methods including majority suppression, authority suppression, and causal logic chains to achieve a high hit rate with a single text.
  • Evaluated ToxicRAG's attack performance on three datasets across four Large Language Models (LLMs), demonstrating superior results compared to state-of-the-art (SOTA) attack models.
arxiv / Preprint
RAGForensics

GHOSTCITE: A Large-Scale Analysis of Citation Validity in the Age of Large Language Models

Co-author 2026.01 – 2026.02

  • Develop GHOSTCITE, an open-source framework that verifies citations at scale, providing a replicable implementation example for venues, publishers, and future work.
  • We evaluate 13 LLMs across 40 domains on 375,440 citations, showing hallucination rates of 14.23–94.93% with significant research domain sensitivity.
  • We analyze 2.2M citations from 56,381 papers (2020–2025), finding 604/56,381 (1.07%) with invalid citations, an 80.9% increase in 2025, and evidence of propagation (up to 16 repeated errors).
WWW 2025
RAGForensics

Traceback of Poisoning Attacks to Retrieval-Augmented Generation

Team Member 2024.09 – 2024.12

  • This paper was completed during an internship in the research group, focusing on the research of retrieval-augmented generation (RAG) system poisoning attacks, defenses, and traceability.
  • Assisted in designing the core RAGForensics system, which employs an iterative process to remove contaminated text and ensures that the final retrieved text remains benign.
  • Designed experiments using datasets such as NQ, MS-MARCO and HotpotQA, established evaluation metrics, and tested the traceability effectiveness of the RAGForensics system.

💰 Scholarships

  • 2025.11: 💰Enlai Zhou Scholarship(￥10000, Top 10), Nankai University(周恩来奖学金初评全校第二，南开大学最高荣誉)；

  • 2025.11: 💰BYD Scholarship(￥10000, Top 0.6%), Nankai University(比亚迪奖学金)；

  • 2025.09: 💰National Scholarship(￥10000, Top 0.6%), China(国家奖学金, 全学院综合排名第一)；

  • 2025.04: 💰First Prize of Yun’an Scholarship(￥5000, Top 1.6%), Nankai University(南开大学“云安密码创新奖学金”一等奖)；

  • 2024.09: 💰National Scholarship(￥10000, Top 0.6%), China(国家奖学金, 全学院综合排名第一)；

  • 2023.10: 💰Academic Excellence Scholarship(￥2000), Nankai University(南开大学学业优秀奖学金)；

  • 2023.10: 💰Literary and Sports Scholarship(￥2000), Nankai University(南开大学文体奖学金)；

🏆 Competitions

  • 2026.04: 🏆The 2nd price of CCSSSC,China(第九届中国高校智能机器人创意大赛-软件系统安全赛华北赛区二等奖)；

  • 2025.12: 🏆The 3rd price of 2025 Haihe Cup(￥5000),Tianjin(天津市“海河工匠杯”技能大赛决赛三等奖)；

  • 2025.10: 🏆Changcheng Cup, China(2025年第五届“长城杯”网络安全大赛“安全卫士”称号)；

  • 2025.10: 🏆The 1st prize of 2025 Beijing–Tianjin–Hebei Security Cup, Tianjin(2025年京津冀大学生信息安全网络攻防大赛决赛一等奖)；

  • 2025.08: 🏆The 1st prize of National Crypto-math Challenge(￥10000, Top 4), China(第十届全国高校密码数学挑战赛全国总决赛一等奖，单赛道第一名，并作为唯一本科生参与特等奖答辩)；

  • 2025.08: 🏆The 2nd prize of National AI Security Contest for College Students, China(第一届全国大学生人工智能安全竞赛作品赛二等奖)；

  • 2025.07: 🏆The 1st prize of National Crypto-math Challenge(rank 1st), Huabei Region(第十届全国高校密码数学挑战赛华北赛区一等奖，单赛道第一名)；

  • 2025.03: 🏆The 3rd Prize (CTF) of National College Student information Security Contest, China(第十八届全国大学生信息安全竞赛创新实践能力赛半决赛三等奖)；

  • 2024.11: 🏆Qiangwang Pioneer Prize of Qiangwang Bei, China(第八届强网杯“强网先锋”荣誉称号)；

  • 2024.11: 🏆Wangding Star of Wangding Bei, China(第四届网鼎杯“网鼎之星”荣誉称号)；

  • 2024.09: 🏆The 1st prize of 2024 Beijing–Tianjin–Hebei Security Cup, Tianjin(2024年京津冀大学生信息安全网络攻防大赛决赛一等奖)；

  • 2023.12: 🏆Third prize of the 15th Chinese Mathematics Competitions, China(第十五届全国大学生数学竞赛三等奖)；

🎖️ Honors and Awards

  • 2026.06: 🎖Outstanding Graduate，Nankai University(南开大学优秀毕业生)；

  • 2026.05: 🎖Outstanding Student Party Member Model, Nankai University(南开大学优秀学生党员标兵，全校仅10人入选)；

  • 2026.05: 🎖Outstanding Youth League member, Nankai University(南开大学优秀共青团员)；

  • 2026.01: 🎉Participated in the Third Information Security Youth Training Camp organized by Tsinghua University and ranked top 20%(参与清华大学网络科学与网络空间研究院组织的第三届信息安全青训营，在考核中排名前20%);

  • 2026.01: 🎖Intermediate Network and Information Security Administrator, Tianjin(晋升网络与信息安全管理员中级工职称);

  • 2025.11: 🎖Excellent student, Tianjin(天津市优秀学生)；

  • 2025.11: 🎖Outstanding Student, Nankai University(南开大学三好学生)；

  • 2025.09: 🎖Excellent Teaching Assistant, Nankai University(南开大学本科课程优秀助教)；

  • 2025.06: 🎖Awarded the Honorary Title of “Top Ten Micro Party Lectures“，Nankai University(南开大学“十佳微党课”荣誉称号)；

  • 2025.06: 🎖Outstanding Team in the National College Student Yimeng Spirit Volunteer Publicity Campaign，China(全国大学生沂蒙精神志愿宣讲活动优秀团队)；

  • 2025.06: 🎖Head of Red Flag Party Branch，Nankai University(南开大学红旗学生党支部主要负责人)；

  • 2025.05: 🎖Outstanding Communist Youth League Cadre, Nankai University(南开大学优秀共青团干部)；

  • 2025.04: 🏆Top 10 Singers in the Second Singing Competition of the College of Cryptology and Cyber Science, Nankai University(学院第二届十大歌手)；

  • 2025.03: 🎖National Information Security Test Program(Grade 1), China(国家信息安全水平考试一级)；

  • 2025.02: 🎖Awarded the Title of “Top Ten Volunteer Service Projects”，Nankai University(南开大学“十佳志愿服务项目”称号);

  • 2024.12: 🎖Sangfor Certified Security Associate Service, China(深信服SCSA认证)；

  • 2024.11: 🎖Outstanding Student, Nankai University(南开大学三好学生)；

  • 2024.10: 🎖Qianxin Certified Emergency Response Engineer, China(奇安信应急响应工程师认证)；

  • 2024.09: 🎖Outstanding Core Member of the “Hao Mi” Group, Nankai University(“豪密”爱国主义宣讲团优秀骨干)；

  • 2024.05: 🎖Outstanding Youth League member, Nankai University(南开大学优秀共青团员)；

  • 2023.10: 🏆Top 10 Singers in the First Singing Competition of the College of Cryptology and Cyber Science, Nankai University(学院第一届十大歌手)；

  • 2023.05: 🎖Third Prize and Outstanding Organization Award in “May Flowers” Choir Competition(as Team Leader), Nankai University(作为学院领队，获南开大学“五月的鲜花”合唱比赛三等奖、优秀组织奖)；

⚙️ Projects

Zipf
Zipf

Revisiting Zipf ’s Law in Passwords: A Taylor Series Modeling Approach

Team Member 2025.04 – 2025.08

  • Reproduced CDF-Zipf from the original paper and conducted multiple rounds of experiments to test the fitting rate, obtaining error results consistent with those in the paper. Optimized the original paper's approach using methods such as the binsort algorithm, achieving a performance improvement of 1.97-3.93 times.
  • Proposed a new model, Exp-Cutoff, and a hybrid model, Mix_CDF, which integrates the strengths of CDF-Zipf and Stretched-exponential. Both models achieved the best KS distance on all test datasets (Myspace, Rockyou, Dodonew, 000webhost, etc.), with a maximum optimization of 55.39%.
  • Conducted subset stability validation experiments by fitting data subsets of varying proportions (10%, 25%, etc.). The resulting errors were all below 5%, demonstrating the stability of the models.
CodeShielder
CodeShielder

CodeShielder: An Expert Vulnerability Detection System Based on Fine-Tuned Large Language Models

Team Leader 2025.03 – 2025.08

  • Built a vulnerability detection system based on fine-tuned large language models and constructed the VulExpert, which outperforms prior works (e.g., VulLLM) in vulnerability detection.
  • In dataset construction, commits were crawled from GitHub, and the VulExpert dataset was built through a series of steps including binary classification, patch crawling, and diff code extraction.
  • For model fine-tuning, the acquired dataset was used to fine-tune the base model Qwen-14B via Q-LoRA, resulting in the fine-tuned model.
  • Fine-tuning tests were conducted on datasets such as BigVul, Draper, DiverseVul and Reveal. The results show that our VulExpert achieves significantly better fine-tuning performance compared to baselines.
Vulnerability Wiki
Vulnerability Wiki

Vulnerability Wiki: a Comprehensive, Systematic and Multi-dimensional Knowledge-sharing Community for Vulnerabilities

Team Member 2025.08 – 2025.10

  • Systematically organize the principles, attack techniques, detection, and defense methods of various vulnerabilities.
  • Track high-risk CVE cases, combining reproduction environments and exploitation chain analysis Provide learning materials tailored to different technical skill levels.
  • Pool community efforts to collaboratively build an open and shared vulnerability encyclopedia.

📖 Educations

💬 Academic Services

  • 2024.09-2025.08, President of the Student Union of the College of Cryptology and Cyber Science, Nankai University(计算机学院和密码与网络空间安全学院学生会主席);

  • 2024.09-2025.08, Vice Captain of “Haomi”, Nankai University(南开大学“豪密”爱国主义宣讲团副团长);

  • 2023.09-2026.06, Class Monitor of the College of Cryptology and Cyber Science, Nankai University(班长);

  • 2025.03-2026.01, Part-time Instructor, Nankai University(南开大学兼职辅导员);

  • 2025.08-2026.06, Organization Committee Member of the Party Branch of the College of Cryptology and Cyber Science, Nankai University(党支部组织委员);

  • 2025.08-2026.06, Vice President of the Information Security Association, Nankai University(南开大学信息安全协会副会长);

  • 2023.09-2024.01, Teaching Assistant for Higher Mathematics Class AI, Nankai University(高等数学AI课程助教);

  • 2024.03-2024.07, Teaching Assistant for Higher Mathematics Class AII, Nankai University(高等数学AII课程助教);

  • 2024.09-2025.01, Teaching Assistant for Higher Mathematics Class AI, Nankai University(高等数学AI课程助教);

  • 2025.03-2025.07, Teaching Assistant for Cryptology, Nankai University(密码学基础课程助教);

  • 2025.09-2026.01, Teaching Assistant for Higher Mathematics Class AI, Nankai University(高等数学AI课程助教);

💻 Internships

  • 2024.07 - 2024.08, 360 Safety Company(360安全集团)

    - Conducted penetration testing and vulnerability discovery tasks, reproduced various types of vulnerabilities, and developed PoC exploits.

NKU Logo PKU Logo