Answers are grounded in this paper and cite their sources.
Demonstrates effective intrusion detection in cloud computing, highlighting improved privacy and accuracy.
Cloud computing environments are increasingly vulnerable to sophisticated cyber threats due to their distributed and dynamic nature. Traditional centralized intrusion detection systems suffer from scalability and privacy limitations, making them less effective in modern cloud infrastructures. To address these challenges, a Federated Learning-based Long Short-Term Memory (Fed-LSTM) framework is proposed for distributed and privacy-preserving intrusion detection. The framework enables collaborative model training across multiple nodes without sharing raw data, thereby ensuring data confidentiality. The model is evaluated using benchmark datasets including CICIDS2017, NSL-KDD, and UNSW-NB15, which are distributed across federated nodes to capture heterogeneous network traffic patterns. Data preprocessing is performed using Min-Max normalization and entropy-based feature selection to improve efficiency and reduce dimensionality. Privacy is ensured through Differential Privacy and Secure Aggregation, while model aggregation is performed using Federated Averaging. Experimental results show that the proposed Fed-LSTM model achieves 95.1% detection accuracy, outperforming traditional machine learning and deep learning baseline methods in terms of accuracy and stability. The proposed approach enhances intrusion detection performance while preserving data privacy, making it suitable for deployment in distributed and dynamic cloud environments.
No takes yet. Share an insight, caveat, or question.
Induru et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: