Hybrid Optimization Machine Learning Framework for Enhancing Trust and Security in Cloud Network

Abstract

The rapidly evolving field of cloud-based data sharing faces critical challenges in ensuring comprehensive privacy protection and trust for both data producers and seekers. Current methodologies often fall short, addressing only one facet of these issues—either privacy or trust—while also struggling with inefficient resource allocation in dynamic cloud environments. This research introduces a novel hybrid paradigm that addresses critical challenges in cloud-based data sharing by integrating enhanced privacy protection, trust generation, and optimized resource allocation. The framework utilizes k-anonymity to protect the privacy of both data producers and seekers by anonymizing participants, thus reducing the risk of data re-identification. For resource allocation, the framework employs the Time-aware modified best fit decreasing (T-MBFD) algorithm, which adapts to fluctuating workloads. Key input parameters for T-MBFD include available resources, job size, and time constraints, while output parameters focus on optimized resource distribution and minimizing wastage. To strengthen trust evaluation, the framework combines an optimized Levenberg algorithm with the Firefly Algorithm. This integration uses trustworthiness scores and service performance as input parameters, resulting in improved accuracy and reliability in trust assessments. Simulation results on a dataset of 95,000 records demonstrate the frameworks robustness and scalability. The proposed framework achieved an average accuracy of 96.416%, an F-measure of 0.976, a precision of 0.958, and a recall of 0.989. Moreover, the framework recorded an AUC-ROC value of 0.990 and specificity of 0.960, with an execution time of 0.22 seconds. These metrics indicate superior performance in privacy protection and trust generation compared to existing approaches, which typically report lower accuracy and trust evaluation metrics. Benchmarking against current models shows consistently higher accuracy, precision, recall, and F-measure values, underscoring the frameworks enhanced capability to safeguard data privacy and accurately assess trust. This study significantly advances secure cloud-based data sharing by offering a robust solution for privacy preservation, trust enhancement, and resource efficiency.

Keywords

• Cloud network security
• data privacy
• hybrid optimization
• k-anonymity
• machine learning
• trust-based mechanisms