Deep Learning–based IDS framework for Cloud Data Security

Abstract

Cloud computing's scalable, adaptable, and cost-efficient processing and storage capabilities have completely transformed big data management. However, moving sensitive data into dispersed cloud environments presents significant privacy and security issues, such as advanced persistent threats, data breaches, and illegal access. This paper proposed an integrated framework that integrates access control and a hybrid deep learning and machine learning-based intrusion detection system (IDS) for real-time threat detection to improve cloud data security. The framework includes data preprocessing, advanced SMOTE for balancing classes, and hybrid feature learning using 1D Residual Autoencoders (RAE) and Convolutional Neural Networks (CNN).The experimental set up is processed on benchmark datasets (UNSB-NB15, NSL-KDD, and WSN-DS), which shows that the model can successfully detect both frequent and rare attacks with an accuracy of 97.6%, 95% and 95.68%, respectively.The integration of intelligent intrusion detection with cloud data security ensures multi-layered security, increasing confidentiality, integrity and availability across cloud infrastructure. Future research will focus on integrating federated learning and blockchain-based trust management to enable decentralized, privacy-preserving, and adaptable cloud security solutions.