Design and Implement a Novel Topology Inference Attack on Decentralized Federated Learning

Federated Learning (FL) is a cutting-edge approach to decentralized Machine Learning (ML) where multiple entities collaborate to train a global model without sharing raw data. Since only the model is shared, FL protects the privacy of the user. However, the training of ML models leaves footprints, so an attacker is able to restore the training information through these tracks. In Decentralized FL (DFL), network topology plays a pivotal role in the overall architecture and security of the system [1]. Understanding and safeguarding the network's topology is essential for ensuring the privacy of data and mitigating security risks in DFL [2]. This project proposal aims to investigate and enhance the privacy and security aspects of DFL by focusing on network topology analysis and the prevention of inference attacks on the network structure.

This project aims to design and implement a topology inference attack which is able to use the training information to recover the topology information for a DFL system. The core objectives of this project question are as follows: 1) to explore the feasibility of designing and implementing topology inference attacks on DFL platform [3, 4]; 2) to assess the impact of these attacks on network privacy, security, and data leakage; 3) to investigate and develop effective defense mechanisms to safeguard DFL systems from such attacks.

[1] Beltrán, E. T. M., Pérez, M. Q., Sánchez, P. M. S., Bernal, S. L., Bovet, G., Pérez, M. G., ... & Celdrán, A. H. (2022). Decentralized federated learning: Fundamentals, state-of-the-art, frameworks, trends, and challenges. arXiv preprint arXiv:2211.08413.

[2] Feng, C., Celdrán, A. H., Vuong, M., Bovet, G., & Stiller, B. (2023). Voyager: MTD-Based Aggregation Protocol for Mitigating Poisoning Attacks on DFL. . arXiv preprint arXiv:2310.08739.

[3] https://github.com/enriquetomasmb/fedstellar

[4] Beltrán, E. T. M., Gómez, Á. L. P., Feng, C., Sánchez, P. M. S., Bernal, S. L., Bovet, G., ... & Celdrán, A. H. (2023). Fedstellar: A Platform for Decentralized Federated Learning. arXiv preprint arXiv:2306.09750.