Room 3006
Cheng Yu Tung Building
Dongdong She
Assistant Professor
Department of Computer Science and Engineering
Hong Kong University of Science and Technology
dongdong@cse.ust.hk
Drop me an email if you are interested in working with me!
I received my Ph.D. from the Department of Computer Science at Columbia University, working with Suman Jana and Baishakhi Ray. Before coming to Columbia, I worked with Zhiyun Qian on Android security at the University of California, Riverside. I earned my bachelor’s degree at Huazhong University of Science and Technology.
I am broadly interested in the intersection of security and machine learning. In particular, I study the safety and reliability of large language models (LLMs) and the security of LLM-based systems, and I explore how LLMs can perform cybersecurity tasks autonomously. My work has earned Distinguished Paper Awards at ASE, IEEE S&P, and ISSTA, the ACM CCS Best Paper Runner-Up Award, and finalist honors in the NYU CSAW Applied Research Competition. Current research projects include:
- LLM Safety: investigating how internal failure modes, such as hallucination, reward hacking, and misalignment, affect the reliability of LLMs and LLM-based agents.
- LLM Security: investigating how external adversaries can compromise LLM-based systems through attacks such as prompt injection and memory poisoning, especially in long-horizon tasks.
- LLM Cyber: exploring the capabilities of LLMs in autonomous cybersecurity tasks, including vulnerability discovery, PoC generation, and exploit generation.
I am fortunate to work with these talented students:
- Yuchong Xie (2024-Fall PhD, BS, MS@SJTU, ex-SJTU 0ops team, intern: Kimi RL team, Qwen foundation model cyber team) co-advised with Prof. Shuai Wang
- Shuangjie Yao (2024-Fall PhD, BS@SJTU, Zhiyuan College, visiting: Columbia University 2025/12-2026/08, intern: Tencent Hunyuan Team - Qingyun Talent Program)
- Zesen Liu (2025-Fall PhD, BS@Xidian U, intern: Tencent Hunyuan Team - Qingyun Talent Program)
- Kaikai Zhang (2025-Fall PhD, BS@SJTU, intern: Qwen foundation model cyber team)
- Zhixiang Zhang (2026-Fall PhD, BS@Fudan U)
- Zihan Zhang (2027-Fall PhD, BS@SUSTech, intern: Qwen foundation model cyber team)
Publications
2027
- ICSEProtecting Agentic Software Intellectual Property via Action-Sequence WatermarkingIEEE/ACM International Conference on Software Engineering (ICSE 2027)
2026
- NeurIPS WSCost-Efficient Theorem Proving via Agent Orchestration in Program VerificationNeurIPS 2026 Workshop on AI for Verifiable Coding, 2026.
- arXivCheap to Hypothesize, Costly to Verify: The Defense Surface of Agentic Vulnerability DiscoveryarXiv preprint
- arXivCompositional Safety Failures in Harness Evolution: Identification and Runtime MonitoringarXiv preprint
- arXivSafe to Check, Unsafe to Use: Relinking at the Compression Boundary of LLM AgentsarXiv preprint
- ASEVerifier-in-the-Loop LLM Solving of Heap Constraints for Concolic ExecutionIEEE/ACM International Conference on Automated Software Engineering (ASE 2026)
- arXiv
- arXiv
2025
2024
2023
- IEEE S&PPrecise Detection of Kernel Data Races with Probabilistic Lockset AnalysisIEEE Symposium on Security and Privacy (S&P 2023)
2022
2021
- SecurityFine Grained Dataflow Tracking with Proximal GradientsUSENIX Security Symposium (Security 2021)
2020
- IEEE S&PNeutaint: Efficient Dynamic Taint Analysis with Neural NetworksIEEE Symposium on Security and Privacy (S&P 2020)
2019
2016
- CCSAndroid ION Hazard: The Curse of Customizable Memory Management SystemACM SIGSAC Conference on Computer and Communications Security (CCS 2016)
2015
- CCSAndroid Root and Its Providers: A Double-Edged SwordACM SIGSAC Conference on Computer and Communications Security (CCS 2015)