Trust and Security Enhancement in Cloud Networks Using a Multiscale-Dilated Depthwise Separable Convolutional Network

In cloud networks, maximum resource utilization, high trust, and security remain basic issues. To overcome such inherent drawbacks, this article proposes a novel deep learning framework to optimize trust and security in cloud networks. Initially, hierarchical attribute-based encryption is employed here to encrypt personal data and hide users' locations and reputations, thereby effectively preventing re-identification attacks. For resource allocation, a multiscale-dilated depthwise separable convolutional network is employed, which dynamically adapts to varied workloads and user request demands. Also, the model leverages key input parameters, such as available resources, job size, and time limit, to generate optimally adjusted resource distribution plans that avoid resource waste and ensure high service efficiency. To further enhance the trustworthiness of cloud services, an orangutan optimization algorithm-based trust assessment module is incorporated. It considers both trust scores and service performance metrics as input parameters to evaluate trustworthiness, thereby supporting more accurate and responsive trust judgments among nodes in a distributed network. The proposed framework is experimentally tested on the Integrated Cloud Security 3Datasets, which includes real-world cloud attack and performance information. The suggested framework not only enhances cloud operational transparency and security but also provides a robust, stable cloud environment. The model is implemented in Python and evaluates performance metrics including accuracy, recall, F-measure, precision, execution time, and mean absolute error. It demonstrates improved accuracy, improved precision, and lower execution time than other comparable schemes.