A Hybrid BiLSTM-Attention and Homomorphic Encryption Framework for Privacy-Preserving Threat Detection in Cloud Computing Environments
- 1 October 2025
- journal article
- Published by World Scientific Pub Co Pte Ltd in SPIN
- Vol. 15 (05)
Abstract
Cloud computing has become indispensable for modern digital services in healthcare, finance and education, yet its distributed and multi-tenant nature exposes it to increasingly sophisticated cyberattacks. With the anticipated integration of quantum computing into cloud infrastructures, the need for robust, privacy-preserving and fault-tolerant security mechanisms is greater than ever. Traditional threat detection models, including Convolutional Neural Networks, Support Vector Machines and Autoencoders, rely on raw or partially anonymized data, which risks user privacy and lacks the ability to effectively capture sequential dependencies in dynamic cloud traffic. To address these limitations, we propose a Hybrid BiLSTM-Attention and Homomorphic Encryption Framework for intrusion detection in cloud environments. The BiLSTM component extracts bidirectional temporal dependencies from encrypted traffic, while the attention mechanism highlights critical time steps to improve detection accuracy. Homomorphic Encryption ensures sensitive data remains encrypted throughout the detection pipeline, supporting privacy without sacrificing computational performance. Experimental results on a benchmark intrusion detection dataset demonstrate superior performance, achieving 99.28% accuracy, 99.26% precision, 99.89% recall and a 99.5% F1-score, along with a low false-negative rate (0.1%) and high AUC (0.9847). Training (1.2[ ]s) and prediction (0.2[ ]s) times confirm the framework’s suitability for real-time deployment in privacy-critical applications. Importantly, the proposed system embodies principles of fault tolerance by maintaining resilience against adversarial interference and data leakage, a requirement that closely parallels the reliability demands of emerging quantum-classical hybrid systems. This synergy between deep learning, encryption and fault tolerance provides a pathway toward secure and privacy-preserving cloud infrastructures in the quantum computing era.Keywords
This publication has 37 references indexed in Scilit:
- EPAD: empowering privacy-preserving anomaly detection through homomorphic encryption and machine learningCluster Computing, 2025
- PriCollabAnalysis: privacy-preserving healthcare collaborative analysis on blockchain using homomorphic encryption and secure multiparty computationCluster Computing, 2025
- Approximate homomorphic encryption based privacy-preserving machine learning: a surveyArtificial Intelligence Review, 2025
- Homomorphic Encryption and Collaborative Machine Learning for Secure Healthcare AnalyticsSecurity and Privacy, 2024
- Comprehensive framework for implementing blockchain-enabled federated learning and full homomorphic encryption for chatbot security systemCluster Computing, 2024
- Random forest evaluation using multi-key homomorphic encryption and lookup tablesInternational Journal of Information Security, 2024
- EVAD: encrypted vibrational anomaly detection with homomorphic encryptionNeural Computing and Applications, 2024
- A 2D Cryptographic Hash Function Incorporating Homomorphic Encryption for Secure Digital SignaturesAdvanced Materials, 2024
- Privacy‐preserving credit risk analysis based on homomorphic encryption aware logistic regression in the cloudSecurity and Privacy, 2024
- A Fully Privacy-Preserving Solution for Anomaly Detection in IoT using Federated Learning and Homomorphic EncryptionInformation Systems Frontiers, 2023