In the era of rapid technological advancement, many organizations shift their data to the cloud. At the same time, valuable data should be made available to various stakeholders when required for analysis, storage, and data consumption. Conversely, to attain security and preserve privacy while data is present on the cloud, data sharing involves significant challenges. In this paper, the authors have integrated cryptography technologies with probabilistic frameworks and machine learning (ML) to introduce a unique model that assists in secure data distribution. The model describes the rules and protocol based on which data can be allocated to various parties from the cloud. The proposed HyMLCPP (Hybrid Machine Learning Cryptography Predictive Probabilistic) model facilitates minimizing the computational time in digital image encryption. Further, it enhances data distribution and sharing practices among users while safeguarding privacy by integrating differential privacy techniques and ML models. The proposed model minimizes the risk by simultaneously detecting vulnerabilities in the cloud-IoT environment. The experimental outcomes demonstrate the proficiency of the proposed model over different self-generated and standard data sets. The proposed model was tested by the authors on standard datasets such as the CICIDS 2017 Dataset, Kaggle Dataset, AIND DDoS Dataset , and CTU-13 Dataset for DDoS attack detection. The proposed model shows accuracy and precision values of about 100% and 99.98% respectively, for standard and self-generated data sets. The proposed HyMLCPP model was tested using various accuracy measurement parameters such as F1 score, recall value , and precision score . The probabilistic features of the proposed model tend to identify the future events as a DDoS assault, which further makes the HyMLCPP model unique and stronger in comparison to the prior works that have proved the effectiveness.
Pathak et al. (Sun,) studied this question.