About
Abdenour Soubih
/ About
I'm a PhD student at InfoLab, in the Department of AI Systems Engineering at Sungkyunkwan University (SKKU), South Korea. My research sits at the intersection of machine learning security, federated learning, and the emerging space of agentic AI systems built on top of large language models.
I work on understanding how ML systems behave under realistic threat models — poisoning, inference, heterogeneity — and on designing mechanisms that keep them robust, private, and accountable when they're deployed beyond clean benchmark conditions. Alongside research, I'm a graduate teaching assistant for undergraduate C programming, and I maintain my course materials on this site.
/ Research areas
Security & Adversarial ML
Threat models for ML pipelines and defenses that hold under realistic deployment.
Federated Learning
Robustness under non-IID clients, poisoning attacks, and privacy leakage.
Agentic AI Systems
Multi-agent coordination, tool use, and safety properties of LLM-driven agents.
Large Language Models
Evaluation, alignment, and trust properties of LLM-based applications.
Multi-agent Systems
Communication and emergent behavior across coordinating agents.
Trustworthy AI
Privacy, robustness, and auditability as first-class system properties.
/ Publications
/ Education
/ Experience
Graduate Teaching Assistant — SKKU
Labs and weekly exercises for undergraduate C programming. Course materials live on this site.
WaterMed4.0 — PRIMA Project
Smart irrigation IoT systems with research teams in Spain and Turkey.
Telecom Internship — Access Telecom
Cellular network performance and KPI analysis on real operator data.
IoT & Distributed Systems
FIWARE platform design, NGSI-LD context management, and real-time data pipelines.
/ Toolbox
/ Languages
/ Get in touch
Open to research collaborations, PhD-track opportunities, and ML/AI engineering roles in trustworthy and privacy-aware AI. Reach me at abdenour@skku.edu.