Decentralized Federated Learning for Intrusion Detection in 5G Networks

The exponential growth in 5G network deployments has significantly increased the complexity and volume of network traffic, presenting new security challenges. Traditional centralized security mechanisms struggle with scalability, privacy concerns and the dynamic nature of threats. This thesis addresses these challenges by exploring a novel decentralized approach, combining Federated Learning (FL) [2] and Multi-Agent Systems (MAS) [3], to effectively identify and mitigate potential threats within 5G network traffic.

The objective of this thesis is to build a distributed, scalable and privacy-oriented artificial intelligence model for detecting network intrusions in 5G environments. Specifically, the research focuses on applying the Asynchronous Consensus-based Learning (ACoL) algorithm [4], a decentralized FL method, to train classifiers implemented by intelligent agents. The project comprises three key components: (i) the 5G-NIDD dataset [1], containing the 5G network traffic samples, including attack packets; (ii) the decentralized ACoL algorithm for FL; and (iii) the intelligent agents' topology responsible for independently training classification models and collaboratively sharing learned knowledge.

The student is expected to (i) conduct a thorough literature review on security in 5G networks and decentralized FL methodologies; (ii) develop a prototype of the decentralized FL algorithm for scenarios with non-IID data distributions; and (iii) perform the experiments to evaluate the efficacy and scalability of the developed solution. Experience with Python programming and the PyTorch framework is required. This thesis offers an opportunity to contribute to the scientific community with a potential publication.

Sources to Consider:

[1] Samarakoon, S., Siriwardhana, Y., Porambage, P., Liyanage, M., Chang, S. Y., Kim, J., ... & Ylianttila, M. (2022). 5G-NIDD: A comprehensive network intrusion detection dataset generated over 5G wireless network. arXiv preprint arXiv:2212.01298.

[2] B. McMahan and D. Ramage. "Federated learning: Collaborative machine learning without centralized training data." Google Research Blog, vol. 3, 2017.

[3] M. Wooldridge and N. R. Jennings. "Intelligent agents: Theory and practice." The Knowledge Engineering Review, vol. 10, no. 2, pp. 115–152, 1995.

[4] Carrascosa, C., Pico, A., Matagne, M. M., Rebollo, M., & Rincon, J. A. (2024). Asynchronous consensus for multi-agent systems and its application to federated learning. Engineering Applications of Artificial Intelligence, 135, 108840.

  Supervisors: Nasim Nezhadsistani, Francisco Enguix