Toward Resilient Smart Cities: A Systematic Comparative Survey of Blockchain-Integrated Deep Learning Architectures for IoT Authentication and Secure Communication

Abstract

The growing adoption of Internet of Things (IoT) devices supports the development of smart cities but also brings serious security risks due to device diversity, limited computing resources, and an expanding attack surface. Traditional security tools are falling behind as cyber threats become more advanced, which has led researchers to explore new combinations of technologies. This survey offers a comparative analysis of how blockchain and deep learning (DL) can work together to strengthen IoT security in smart city settings, with particular attention to authentication methods and communication protocols. We look at how blockchain enables decentralized trust, tamper-resistant authentication, and auditable access control through Decentralized Identifiers (DIDs), smart contracts, and lightweight consensus protocols. In parallel, we study how deep learning supports adaptive intrusion detection, risk scoring, and anomaly monitoring using Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Graph Neural Networks (GNNs), and autoencoders. A central contribution of this work is a six-dimensional comparison framework that rates approaches on scalability, latency, security level, energy efficiency, privacy preservation, and interoperability. Our findings show that hybrid blockchain-DL designs outperform single-technology solutions across most of these dimensions, though open problems remain in cross-domain interoperability, consensus overhead, and adversarial robustness of DL models. We also evaluate eight benchmark IoT security datasets and outline a phased research roadmap covering short-term (2025–2027), medium-term (2027–2029), and long-term (2029–2032) goals. This survey is intended to serve as a practical reference for researchers and engineers working on secure, intelligent IoT systems for future smart cities.