A Feature-Aware Federated Learning Framework for Unsupervised Anomaly Detection in 5G Networks
Keywords: Federated Learning, Anomaly Detection, 5G Networks, Privacy-Preserving
Abstract: The expansion of 5G networks has led to remarkable data volume and complexity, introducing significant security challenges that require the implementation of robust and scalable anomaly detection mechanisms. Traditional centralized approaches pose privacy risks and scalability challenges due to the distributed nature of 5G infrastructures. Federated Learning (FL) offers a decentralized solution but often overlooks the importance of feature relevance and privacy preservation during model aggregation. This paper introduces a novel Feature-Aware Federated framework that integrates feature importance into the aggregation process while ensuring differential privacy. We employ integrated gradients to compute feature importance for each client, aggregate them globally with differential privacy noise, and use these insights to weight model parameters during aggregation. Additionally, we propose Dynamic Feature Importance Adaptation (DFIA) to update feature importance occasionally, enhancing the model's adaptability to evolving data distributions. Experimental results demonstrate that our framework outperforms traditional federated approaches like FedAvg and FedProx in unsupervised anomaly detection tasks within 5G networks, achieving higher accuracy and robustness while preserving data privacy.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 14294
Loading