Opportunities

Currently available opportunities

Phd

Sustainable and Robust Federated Learning for Cyber Threat Detection in 5G Networks

Applications are invited for a self-funded, 3-year full-time or 6-year part-time PhD project.

The work on this project will:

• Exploration and application of advanced techniques for preprocessing, includes investigating strategies to collect data and preprocessing 5G network data

• Comprehensive investigation, analysis, and design of Federated Learning Approaches and Algorithms

• Designing approaches that enhance the efficiency, adaptability, robustness, and sustainability of federated learning models, ensuring optimal performance in cyber threat detection.

• Simulation, evaluation, and validation of the developed models. The focus extends to real-world datasets, particularly within the realm of cybersecurity.

Please refer to the link here for more information. 

You can to contact Dr Rahim Taheri ( rahim.taheri@port.ac.uk ) to discuss your interest before you apply.

Master Projects

1- Security Evaluation of Retrieval-Augmented AI Systems: This project investigates prompt injection, unauthorised information disclosure and defensive controls in AI systems that generate answers from external documents. 

2- Robustness of Machine Learning Against Data Poisoning in 5G/ 6G network: This project examines how manipulated training data affects model behaviour and evaluates methods for detecting and reducing poisoning attacks. 

3- Privacy Risks in Machine Learning Models: This project investigates whether model outputs reveal information about training data and assesses approaches to reducing privacy leakage. 

4- Resilience of AI-Based Malicious URL Detection: This project develops and evaluates malicious URL detection models under changing attack patterns, with particular attention to evasion techniques. 

5- Security Controls for AI Chatbots Handling Sensitive Information : This project examines how AI chatbots process sensitive information and evaluates access controls and response filtering against unauthorised disclosure.