New Horizons of Science, Technology and Culture Vol. 12

Digital twin technology has evolved from a conceptual modelling tool into an operational backbone for cyber-physical systems across manufacturing, energy, transport, and healthcare. As these virtual replicas synchronise continuously with physical assets, they extend the attack surface available to adversaries while simultaneously offering a new vantage point for defence. This review synthesises recent literature on the intersection of digital twins, artificial intelligence, and cybersecurity, examining how machine learning, federated learning, and explainable models are being applied to threat intelligence, anomaly detection, and resilience engineering within cyber-physical environments. The review traces the evolution of digital twin security architectures, evaluates zero trust and identity-centric approaches to industrial Internet of Things protection, and considers the role of explainability in building operator trust in automated security decisions. Particular attention is given to the dual-use nature of digital twins, which can serve as honeypots, simulation testbeds, and attack surrogates while themselves becoming targets for model poisoning and data manipulation. The review identifies persistent gaps in standardisation, dataset realism, computational overhead, and regulatory alignment, and outlines future research directions centred on federated and privacy-preserving intelligence sharing, post-quantum readiness, and human-centred trust calibration. The discussion is intended to support researchers, system architects, and policymakers seeking a coherent account of how digital twins are reshaping the practice of cybersecurity in cyber-physical infrastructure.