I’m Marzia Tahsin, an undergraduate student of Computer Science at the University of Maryland, Baltimore County (UMBC), affiliated with the Center for Women in Technology (CWIT). I’m currently a senior pursuing a Bachelor’s degree, graduating in December 2026, and I work as a TA for CMSC 201.

My research interests lie in federated learning, differential privacy, and applied AI/ML across healthcare and cybersecurity. I’m currently co-authoring a study on frequency-aware differential privacy for federated Alzheimer’s MRI staging, and I’m part of an ongoing research collaboration through INSuRE (Information Security Research and Education), a multi-institutional cybersecurity research consortium backed by the NSA, working with NSA-affiliated mentors on behavioral AI for Kubernetes security.

Research

• Co-authoring Frequency-Aware Differential Privacy for Federated Alzheimer’s MRI Staging: Protecting Early-Stage Diagnostic Signals (manuscript in preparation), which examines how differential privacy constraints erode diagnostic accuracy at each Alzheimer’s disease stage in a federated Swin-Transformer model, and introduces a frequency-aware noise allocation strategy to protect early-stage detection signals.
• Collaborating with a UMBC research team through INSuRE and NSA mentors on Behavioral AI for Kubernetes Security, building a telemetry pipeline (Prometheus, Falco, eBPF, Grafana) and applying an Isolation Forest model for runtime anomaly detection.
• Completed the AI4ALL Ignite Fellowship, designing an EEG-based emotion recognition system with a Random Forest Classifier, achieving 62.9% accuracy despite class imbalance.