This paper introduces a novel model, MUFD (Multi-view User-based Fake News Detection), which captures various perspectives of user engagement in news dissemination by incorporating propagation graph structures, temporal dynamics, and personal attributes. Graph structure, temporal structure, and personal attributes. Unlike previous studies that use the final propagation structure, our user dynamic propagation graph structure model leverages differences in propagation structures across time windows to learn dynamic changes over time. Additionally, we construct a user propagation temporal change model to capture the temporal order of user participation and a user personal attribute model to learn inherent user attributes. These models are integrated into a unified framework, effectively learning the multifaceted characteristics of users involved in fake news dissemination. Our results show that using multiple views is effective in detecting fake news, even without using news and comments contents. Experiments on real-world datasets demonstrate that our framework significantly improves fake news detection performance compared to state-of-the-art baselines and proves its capability to detect fake news earlier than existing methods.