AI-POWERED DATA MASKING FOR PRIVACYPRESERVING CLOUD DATA SHARING

Abstract

Cloud computing has become increasingly dependent on using cloud environments for storing and processing of sensitive data and thus ensuring data privacy in cloud environments is very important. In this research, we introduce an AI powered data masking framework that covers all levels of precision in order to improve the privacy preserving data sharing in the cloud environment. The proposed methodology use advanced machine learning algorithms e.g Generative Adversarial Networks, GANs to generate synthetic data that resembles the original datasets ensuring secure data exchange without disclosing sensitive data in the datasets. Moreover, data leakage is prevented by integrating differential privacy techniques to provide mathematical guarantees of data leakage even in the presence of malicious entities trying to reverse engineer masked data. The framework adopts federated learning models to do decentralized model training, in which the risk of exposure to direct data is reduced. We evaluate it and show good data utility and acceptable performance overheads lending itself to scalable cloud architectures. Adoption of light weight encryption algorithm to operate efficiently in resource constrained environments including the IoT devices. Additionally, the solution also focuses on employing secure multi party computation (SMPC) to keep the data encrypted through the entire computation process. The experimental results demonstrate the reduction of risks due to the data exposure, in comparison to the traditional anonymization techniques, while still providing utility for analytics tasks. In this regard, this research helps address the growing need for privacy preserving AI solutions, which is consonant with the latest legislations including GDPR and HIPAA.