Abstract: Residential neighborhoods are fundamental units for analyzing urban safety resilience and its spatial differentiation. Existing studies rely largely on static data for macro-scale assessment and therefore struggle to capture the dynamic evolution of urban resilience at fine spatial scales. This study uses residents' spatiotemporal mobility disruption during disasters as the core indicator of resilience and develops a data-driven framework for assessing pluvial flood resilience at the neighborhood scale. Based on mobile phone signaling data, the framework is applied to 1,087 residential neighborhoods in Shenzhen during the concentrated rainfall period of the 2023 "9.7" extreme rainstorm. The results show that commodity-housing neighborhoods have significantly higher resilience than urban villages, while neighborhoods built in different periods display clear heterogeneity in both disaster-time and post- disaster resilience. Correlation analysis confirms that road waterlogging is an important driver of residents' travel disruption and validates the effectiveness of the framework in identifying spatial differences in neighborhood resilience. Demographic validation further reveals that applying mobile phone signaling data to micro-scale resilience assessment may underestimate the real risks faced by aging neighborhoods. The proposed framework provides quantitative support for dynamic disaster monitoring and early warning, urban renewal diagnostics, and sustainable planning.