Research

AI Security and Trustworthiness

• Adversarial robustness of Graph Neural Networks and Large Language Models in cybersecurity, software security, and trustworthy AI applications.
• Security of agentic AI systems, including delegated privileges, cross-application interactions, and confused-deputy behavior.
• Internet, IoT, and critical-infrastructure security.

LLM and GNN Methods

• Prompt optimization and engineering.
• Graph- and semantics-based reasoning for security analysis.
• Trustworthy use of LLMs and GNNs in applied systems.

Application Domains

• Secure and functional source-code generation.
• Hardware design automation.
• Deepfake detection, multimodal deepfake detection, and hallucination detection.
• Internet and malicious-domain security.
• IoT firmware assurance for critical infrastructure.

Earlier Research Foundations

My earlier work focused on machine learning for communications, including channel estimation, spectrum sensing and occupancy prediction, wireless security, indoor positioning, and related signal-processing problems. These foundations continue to inform my work on graph learning, systems, and trustworthy AI.

Research Collaborators and International Connections

Selected collaborators and partner institutions include:

New Jersey Institute of Technology (NJIT)
United States

George Mason University
United States

Rochester Institute of Technology
United States

University of Tennessee at Chattanooga
United States

QCRI, Hamad Bin Khalifa University
Qatar

Alfaisal University
Saudi Arabia

Clarkson University
United States

Istanbul Medipol University
Turkey

Jordan University of Science and Technology
Jordan