A Hybrid Deep Learning Architecture Using Dense Neural Network and Gated Recurrent Unit for Cloud Intrusion Detection

In this paper, we have used a hybrid deep-learning architecture to detect cloud intrusion. The cloud plays an important role in modern data management, accessibility, and storage. It enables organizations to store vast amounts of data remotely. Due to frequent data sharing between platforms and remote access, the cloud environment is more vulnerable to cyber threats. An intruder can access confidential data due to vulnerabilities including inadequate encryption, absence of authentication, poor access control, and so on. We can efficiently use deep learning approaches to enhance cloud security. In this research, we have developed a hybrid dense neural network (DNN) and gated recurrent unit (GRU)-based architecture to detect cloud intrusion. We have also developed individual DNN and GRU-based architectures. In our proposed model, the DNN layers can capture the nonlinear complex pattern, while GRU can identify the temporal dependencies of network traffic data. In this research work, we have utilized the network portion of the ToN-IoT dataset. For multiclass classification, the hybrid DNN-GRU model has achieved the highest performance, obtaining an accuracy of 98.66%, precision of 98.70%, recall of 98.66%, f1 score of 98.67%, and AUC of 99.98%. Moreover, the performance for detecting nine different attacks has also been evaluated. This case shows the hybrid DNN-GRU model performs well, identifying backdoor and ransomware attacks with 100% accuracy. Thus, this hybrid architecture works more robustly and efficiently to detect cloud intrusion by combining the strengths of both DNN and GRU, thereby strengthening cloud security.