Data security and privacy in cloud computing-based internet of things environments: current gaps and emerging solutions

Abstract

The Internet of Things (IoT) is a revolutionary innovation that enables greater automation, efficiency, and ease of use across various domains. Cloud computing is an efficient solution for processing and analyzing the large volume of data generated by IoT components. Data flows from edge devices to cloud environments demand strong end-user authentication, advanced security, strong recovery procedures, and user privacy guarantees. This study presents a comprehensive survey of existing literature on security in cloud-based IoT environments. In addition, associated problems are identified, and practical solutions are suggested. Based on the literature review, shortcomings in current methods of addressing data protection and privacy threats are identified. Ensuring secure data transmission requires significant advances in cloud and IoT device architecture and the development of reliable communication protocols. Emerging technologies, including blockchain, artificial intelligence, fog computing, and edge computing, hold promising prospects. These technologies could mitigate the threats associated with centralized cloud storage. In this context, the leading security threats, categorizations, and reactive plans for protecting data in cloud-based IoT systems are comprehensively reviewed. Blockchain-assisted IoT authentication frameworks have been reported in prior studies to reduce authentication and verification latency by approximately 26–35% compared to centralized authentication schemes and Practical Byzantine Fault Tolerance (PBFT)-based baselines, as observed in simulation-based evaluations under controlled network conditions and specific system configurations. In parallel, machine learning-based intrusion detection systems, particularly deep learning and ensemble models, evaluated on benchmark datasets such as CICIDS2017, TON_IoT, and UNSW-NB15, have achieved classification accuracies exceeding 99% under certain experimental settings, depending on feature selection, model architecture, and dataset-balance constraints.