AI-Driven Cloud Security Framework Using Deep Feature Learning and Hybrid Threat Classification

As the use of digital systems increases, there is a significant concern about protecting sensitive information in the cloud. By combining Artificial Intelligence (AI) and Cloud Computing (CC), it is possible to provide smart and scalable applications to identify and prevent cyber threats. For this reason, this research presents a Secure-intelligence Network (Sec-iNet), an adaptive information security framework which uses security logs and network traffic data from cloud platforms. In the preprocessing phase, data cleaning and normalization are done to remove all noise. In the feature extraction, statistical features and security features are extracted, like IP address frequency, login attempts, and data transferring patterns. To enhance the proposed Sec-iNet's feature learning capability, a Deep Autoencoder and residual blocks combined with DAER-X is proposed for deep feature learning. Now, the significant deep feature set is passed to the classification layer, which is composed of Random Forest (RF) for faster decision making, Support Vector Machine (SVM) for precise boundary-based threat classification, and Deep Spatio-Temporal Neural Networks for capturing complex temporal correlations in network traffic are used to detect unusual behaviour and potential threats. The proposed Sec-iNet enhances the accuracy in threat detection and reduces false positives by considering data privacy as well. Besides, the performance of the proposed model is validated with benchmark models by means of accuracy, which is 5.22%, 6.41%, and 5.48% better than CNN, RF, and SVM in order.