Dr. Amanullah Yasin

PhD Theme/Topic: Privacy-Preserving Federated Soft Biometrics for Real-Time Covert Urban Surveillance

Supervisor: Dr. Amanullah Yasin, Professor
Contact # 0321-8040754
Email: amanyasin.h11@bahria.edu.pk
Campus/School/Dept: BU H-11
RAC Approved Supervisor for Research Areas: Soft Biometrics, Federated Learning, Privacy Preservation, Covert Surveillance, Edge & Cloud Computing, Deep Learning, Crowd Analytics, Activity Recognition, Differential Privacy, Smart City Security.

Supervisory Record:
PhD Produced: 03
PhD Enrolled: 01
MS/MPhil Produced: 50+
MS/MPhil Enrolled: 02

Topic Brief Description:
This PhD research proposes a unified, privacy-preserving soft biometrics framework for covert urban surveillance by integrating federated learning, multi-modal deep learning, and edge-cloud intelligence. Unlike traditional surveillance systems that rely on intrusive biometrics or centralized data processing, the proposed approach leverages non-intrusive soft biometric traits (e.g., gait, clothing, posture, activity patterns) to enable real-time crowd understanding while preserving individual privacy. The framework will employ federated learning to decentralize model training, ensuring that sensitive data remains at the source, combined with edge computing for low-latency inference and cloud-based scalability for large-scale analytics. The research also addresses critical challenges related to data privacy, model robustness, bias, and ethical compliance, aiming to deliver a scalable, secure, and socially acceptable surveillance architecture applicable to smart cities, border security, and critical infrastructure monitoring.

Research Objectives/Deliverables:

• Design a privacy-preserving federated learning framework for soft biometric-based surveillance that avoids centralized data collection.
• Develop and evaluate deep learning models for soft biometric feature extraction (e.g., gait, clothing, posture) in real-world urban scenarios.
• Implement a real-time edge-enabled surveillance pipeline integrating edge inference with cloud-based aggregation for scalable deployment.
• Assess privacy, performance, and ethical trade-offs (accuracy vs. privacy, bias, and robustness) in the proposed system.

Research Questions:

• How can soft biometric features be effectively modeled for reliable and non-intrusive surveillance in urban environments?
• How can federated learning be adapted to handle non-IID, distributed surveillance data while preserving privacy?
• What is the optimal edge–cloud architecture for achieving real-time performance in soft biometric surveillance systems?
• What are the trade-offs between privacy, accuracy, and fairness, and how can they be balanced in such systems?

Candidate’s Eligibility Profile:

• The applicant must have an MS/ MPhil/ Equivalent degree in CS/ SE/ AI/Data Science/ EE/ CE/ Cyber Security / Information Security/ relevant areas with CGPA > 3.0. Besides, applicants must have a strong background in mathematics, optimization theory and related fields.
• Proficiency in Python and relevant ML/DL frameworks (TensorFlow, PyTorch, Keras). Demonstrated knowledge of Computer Vision (object detection, image classification, tracking). Familiarity with deep learning architectures: CNNs, Transformers, YOLO variants. Understanding of biometrics (hard and/or soft) or surveillance systems is a strong plus.
• Candidates should thrive in a team environment and have excellent communication skills to actively contribute to team research efforts.
• Strong academic writing skills in English; ability to present complex technical ideas clearly in written and oral form.
• Availability to participate in collaborative research activities, lab meetings, and potential national/international conference presentations.