Research Focus
What I work on
My research centers on the security and trustworthiness of modern AI systems — from single LLMs to distributed, multi-agent pipelines.
01
LLM Security
Security evaluation, prompt injection, data leakage, adversarial behavior, and secure LLM applications.
02
Multi-Agent AI Security
Trust propagation, agent hijacking, privilege boundaries, authorization, and security of agentic workflows.
03
RAG Security & Privacy
Security evaluation of distributed RAG systems, membership inference, knowledge extraction, poisoning, source manipulation, and availability attacks.
04
Trustworthy AI
Structural authorization, independent verification, least privilege, policy enforcement, and safe execution.
05
AI Privacy
Privacy risks in retrieval and language-model systems and privacy-preserving retrieval approaches.
Research
Research
A detailed look at the problems I investigate, the systems I study, the attacks and threats I evaluate, and the defenses I design — with methodology, findings, and access to the underlying work.
Completed Thesis — BRAC University, 2026
A Security Evaluation Framework for Distributed Retrieval-Augmented Generation Systems
A systematic, seven-step security evaluation framework for distributed RAG systems, deriving six attack classes across the CIA triad and six corresponding defenses — including a previously undocumented attack, Grounding-Farming.
Undergraduate Researcher · Supervised by Dr. Swakkhar Shatabda
Accepted — IEEE CARS 2026 (Main Conference)
Trust Propagation and Structural Containment in Multi-Agent LLM Pipelines
An empirical security study of a four-agent LLM pipeline, examining whether a compromised low-privilege agent can influence a high-privilege executor, and whether structural authorization can contain the attack even when an upstream model judgment is compromised.
Researcher
Interested in working together?
I am open to research discussions, collaborations, and opportunities related to AI security, LLM security, RAG, multi-agent systems, privacy, and trustworthy AI.