1. Multimodal Biometric Fusion for Security and Privacy
Description: This initiative explores biometric systems that integrate facial recognition, voice authentication,
and eye tracking to strengthen security and privacy in digital platforms. Building on work in multimodal biometric fusion, the lab
evaluates deep learning based and fuzzy logic based fusion strategies that enhance robustness against spoofing.
Students gain experience with signal processing, feature engineering, fusion architectures, and performance evaluation, with
applications spanning banking, telehealth, and secure access control.
2. Deep Learning for Emotion Recognition and Human-Robot Interaction
Description: Extending prior work in facial expression analysis and multimodal emotion modeling, this project
investigates how social robots can interpret and respond to human emotions in real time. Convolutional networks and sequence models
are used to integrate facial expressions, speech prosody, and gestures.
The goal is to design socially aware agents that improve engagement and adherence in healthcare assistance, customer support, and
educational settings.
3. Generative Models for Synthetic Data and Anonymization
Description: This project studies generative adversarial networks and related models that can synthesize realistic,
privacy-preserving datasets for healthcare and cybersecurity applications.
The work balances data utility and disclosure risk, providing methodologies for partners who require large-scale datasets while adhering
to regulatory and responsible design requirements.
4. Behavior-Driven Cybersecurity Enforcement
Description: Building on research in policy-based access control, this project examines how behavior-aware models can
continuously adapt security rules based on user actions.
Students analyze login patterns, network activity, and file access logs to detect anomalies and inform automated controls suitable for
enterprise environments.
5. Explainable AI for Trustworthy Decision Support
Description: As AI models influence high-stakes decisions, this project focuses on frameworks that make deep learning
models more interpretable and auditable.
Techniques such as saliency mapping, attention analysis, and post hoc explanation are applied to applications in emotion recognition,
health analytics, and cyber risk assessment to improve user trust and support compliance.
