LLM security
Understanding the security risks and failure modes of large language models, and developing methods to make them safer and more dependable.
Open to research collaboration
Vu Hai Dang
AI Engineer & Researcher
LLM4Code, from the security side — secure code generation, vulnerability detection and safety guardrails for models that write real software.
Scroll to explore
About
Secure and Trustworthy AI for Software Systems
My research lies at the intersection of Software Engineering, Artificial Intelligence, and Cybersecurity, with a particular focus on LLM4Code. I am particularly interested in improving the security, reliability, and trustworthiness of Large Language Models for software engineering tasks, including secure code generation, vulnerability detection, and safety guardrails.
Fields
Software Engineering
Artificial Intelligence
Cybersecurity
LLM4Code
Natural Language Processing
Research
What I work on
Four connected directions around a central question: how can we build more secure, reliable, and trustworthy AI systems for software engineering?
Secure code generation
Improving the ability of language models to generate code that is not only functional, but also secure and robust.
AI for software engineering
Applying AI and language models to support software development, program understanding, vulnerability detection, vulnerability analysis, and code reasoning.
Agentic systems
Building reliable and trustworthy AI agents that can reason, use tools, and interact with complex software environments safely and effectively.
Publications
Selected publications
Walking on Thin Ice? Defending RAG-based Code Recommender Systems Against Poisoning Attacks
Journal of Systems and Software 2026
Towards Knowledge Alignment in Code LLMs: Contrastive Unlearning for Evolving APIs
ICSME 2026 NIER Track Distinguished Paper Award
On the Gap Between Declarative Vulnerability Knowledge and Secure Code Generation in Large Language Models
Information and Software Technology · 2026 Under review
Contact
Get in touch
Let's talk about LLMs and code
I'm open to research collaboration, reviewing, and conversations about LLM security, secure code generation, and retrieval-augmented systems. Email is the fastest way to reach me.