Sep '25

Protecting Federated Learning from Extreme Model Poisoning Attacks via Multidimensional Time Series Anomaly Detection

FLANDERS has been integrated into the FL framework Flower! Here’s the link to the documentation.

Authors: Edoardo Gabrielli, Dimitri Belli, Zoe Matrullo, Vittorio Miori, Gabriele Tolomei

Links:

Bibtex

@article{gabrielli2025flanders,
  author={Gabrielli, Edoardo and Belli, Dimitri and Matrullo, Zoe and Miori, Vittorio and Tolomei, Gabriele},
  journal={IEEE Transactions on Information Forensics and Security}, 
  title={Securing Federated Learning against Extreme Model Poisoning Attacks via Multidimensional Time Series Anomaly Detection on Local Updates}, 
  year={2025},
  volume={},
  number={},
  pages={1-1},
  keywords={Computational modeling;Servers;Data models;Training;Standards;Time series analysis;Predictive models;Anomaly detection;Security;Federated learning;Federated Learning;Robustness;Model Poisoning Attacks;Deep Learning;Anomaly Detection;Security},
  doi={10.1109/TIFS.2025.3608671}}