A Proposed Method for Real-Time Automatic Cloud Storage and Analysis of Detected Suspicious Activities to Ensure Data Integrity and Security

Cloud computing provides flexibility and scalability to enterprises but also poses devastating threats such as data leakage and unauthorized access. This study presents a prototypical security framework of a hybrid cloud system that combines the use of machine learning-based anomaly detection and cryptographic techniques regarding the confidentiality, integrity, and availability (CIA) of dynamic cloud storage to preserve data integrity. The framework uses Random Forest (RF) feature selection and Deep Neural Networks (DNN) to detect anomalies and design a real-time, scalable framework to identify suspicious behavior, supported by a hybrid cryptographic model using Attribute-Based Encryption (ABE) and lightweight algorithms. It is CICIDS2017 and UNSW-NB15 validated and is deployed in AWS and Azure testbeds. Hybrid implemented a high detection rate (98.3 percent) and a low false positive rate (1.2%), and cryptographic operations had an average loading of 4.3 ms/MB. The hybrid model showed better results when compared to standalone RF (95.8%) and DNN (97.6%). Comparisons with other recent articles prove that performance is competitive, although some articles reported accuracy rates above 99.9%. The results establish that the hybrid solution provides high detection, effective cryptography, and great scalability in the context of a multi-cloud environment. Banking, healthcare enterprises and government agencies are examples of enterprises that can implement this framework to limit the risks and realize secure real-time operation. It is possible that future studies can develop quantum-resistant cryptography and federated learning combinations.