The rapid development of smart cities has led to the proliferation of IoT devices and digital twins, which are crucial for urban management, resource allocation, and infrastructure monitoring. However, the integration of these technologies also introduces significant cybersecurity risks, ranging from data breaches to malicious attacks that could disrupt the functioning of critical urban systems (Zhang et al., 2022). This research proposes a cybersecurity framework leveraging soft computing technologies, such as reinforcement learning and deep learning, to secure digital twin environments in smart cities. We begin by collecting diverse IoT and digital twin datasets, which are used to train various ML models aimed at detecting and mitigating cyber threats (Smith & Kumar, 2021). The study emphasizes the application of reinforcement learning for anomaly detection and deep learning for identifying malicious patterns. The proposed framework integrates these models to provide real-time, adaptive security mechanisms, which are tested through simulation experiments. Our findings demonstrate the potential of AI and ML techniques to significantly enhance the cybersecurity of smart cities, offering a scalable and effective solution to counter emerging threats. The framework’s ability to continuously adapt to evolving cyber-attacks makes it a promising tool for future smart city infrastructures (Kumar & Sharma, 2020).