Security and Privacy for Connected Systems: A Layered Review of Cryptography, Intrusion Detection and Privacy-Preserving Learning in IoT, Cloud and Mobile Networks

Abstract

Connected systems now carry medical readings, financial records and personal data across sensors, mobile networks and cloud servers. Each layer of such a system faces different threats. Small IoT devices cannot run heavy encryption, ad hoc networks change their topology all the time, cloud stores hold large volumes of data from many users, and users themselves are targeted by phishing and privacy attacks. Defences are usually designed and evaluated one layer at a time, which makes it hard to see how cryptography, intrusion detection and privacy-preserving learning fit together. This chapter reviews security and privacy mechanisms across four layers: devices and communication links, networks, cloud data stores, and users and applications. It covers lightweight, elliptic curve and quantum-safe cryptography, authentication and blockchain-based trust for constrained devices, security-aware resource management in 5G and 6G networks, moving target defence, learning-based intrusion detection, optimized cloud encryption, phishing detection and privacy protection in federated learning and the metaverse. The review finds that cryptographic designs for constrained devices trade key size and latency against strength, that ensemble and deep learning detectors report very high accuracy on benchmark traffic, and that metaheuristic optimization is increasingly used to tune encryption and detection parameters. Few studies, however, evaluate several layers together or test detectors on attacks that were absent from training. Future work should combine defences across layers, measure their overhead on real devices and evaluate them against new attack types.