|
AUTHOR(S): Sahaaya Arul Mary S. A., Sameer Chauhan, Luv Sachdeva
|
TITLE Improving Intrusion Detection with Federated Learning for Enhanced Privacy Protection |
 PDF |
ABSTRACT There are legitimate privacy and security issues with the processing of massive amounts of sensitive data required to detect breaches, abnormalities, and security risks in network traffic (including IoT). Federated learning, a type of distributed machine learning, lets many people work together to train a single model while keeping data privacy and independence. An alternative to training and assessing the model on a central computer is a federated educational setting, whereby each client learn a local model having the same structure that is trained on its own dataset. After that, an aggregation server receives these local models and uses federated averaging to create an optimal global model. Designing efficient and effective solutions for intrusion detection systems (IDS) is greatly facilitated by this technique. We evaluated the efficacy of federated instruction for IDSs to that of conventional deep learning models in this study. Through the implementation of random client selection, our research shows that federated learning outperformed deep learning in terms of accuracy and loss, especially in data privacy and security-focused situations. We demonstrate via experiments how federated learning may build global models without exposing sensitive data, reducing the dangers of data leaks and breaches. The results show that federated average in federated learning could change the way IDS solutions are made, making them safer, more efficient, and more useful. |
KEYWORDS security of communication networks, federated learning, anomaly detection, intrusion detection systems, and data privacy |
|
Cite this paper Sahaaya Arul Mary S. A., Sameer Chauhan, Luv Sachdeva. (2024) Improving Intrusion Detection with Federated Learning for Enhanced Privacy Protection. International Journal of Communications, 9, 11-22 |
|
