Samir Moustafa

Samir Moustafa is a researcher in the Faculty of Computer Science, affiliated with the Research Group Theory and Applications of Algorithms. His work centers on Graph Neural Networks, adversarial machine learning, and algorithmic security.

He holds a B.Sc. and M.Sc. in Computer Science.

Research Interests
Samir's research explores the vulnerabilities of Graph Neural Networks (GNNs) to adversarial attacks, particularly bit flip attacks, and develops defense frameworks to enhance model robustness. He also investigates the role of information redundancy in GNNs, analyzing the trade-offs between storage overhead and model performance. His work integrates theoretical computer science with practical machine learning applications to improve the security and efficiency of graph-based AI systems.

Publication Trends
Recent publications (2023-2025) demonstrate a focus on securing GNNs against bit flip attacks and understanding redundancy mechanisms. His 2025 paper introduces "Crossfire", an elastic defense framework, while his 2024 works analyze attack indifference via the Weisfeiler-Leman algorithm and the dual nature of redundancy. These contributions highlight his expertise in both offensive and defensive strategies for graph-based machine learning.

Scientific Awards

• Best Paper Award at ECML PKDD 2023 (awarded September 22, 2023)

Advising and Grants
No public information is available regarding graduate student advising or research grants.

Labs and Teams
Samir is an active member of the Research Group Theory and Applications of Algorithms, contributing to collaborative projects on algorithm design and machine learning security.