In recent decades, user communication has been digitalized with some advanced applica- tions. However, securing the digital cloud system is complicated because of the vulner- ability of large files and malicious events. Therefore, a present research study intended to design a novel Dragonfly-based Genetic Deep Belief Network (DGDBN) technique to protect the VM from malware activities in the cloud environment. Hence, to validate the presented model, the cloud user files data was considered and imported to the system as input. Then further processes such as preprocessing feature extraction, attack detection and classification were performed. Once the malicious event is predicted, it is neglected by the cloud user environment. Furthermore, implemented novel DGDBN model is tested in the MATLAB programming environment. Finally, the performance parameters like accuracy, precision, reconfiguration time, Recall, F-measure, and data overhead were measured and compared with associated approaches.