Island tourism destinations integrate both marine and terrestrial attributes, resulting in highly fragile ecosystems that face increasing environmental pressures. Under the combined effects of climate change and intensive tourism development, conventional assessment methods often fail to capture the dynamic evolution of tourism ecological security in these regions. Furthermore, most existing studies depend on static weighting approaches, which overlook inherent ecosystem dynamics and sudden fluctuations, leading to delays in identifying key security drivers and risk-sensitive factors. Most current studies employ statically weighted models that do not adequately account for the intrinsic dynamism and potential for sudden fluctuation in ecosystems. This oversight results in significant delays in the detection of pivotal security drivers and factors sensitive to risk.To address these gaps, this study introduces a novel evaluation framework by integrating variable-weight theory and a grey cloud model within the Capacity-Supporting Capacity-Attractiveness-Sustainability-Development Capacity (CSAED) conceptual system. Applied to the Zhoushan Archipelago, the results indicate that: (1) from 2008 to 2022, the overall ecological security index showed a fluctuating upward trend, increasing from 50.793 to 83.152 and shifting from a "risk" to a "secure" state; (2) Significant differences were observed among subsystems: both supporting capacity and carrying capacity demonstrated steady growth, whereas sustainability and development capacities showed fluctuating trends, characterized by an initial rise followed by a decline; meanwhile, attractiveness decreased gradually throughout the study period. (3) The combination of variable-weight theory and the grey cloud model improves diagnostic sensitivity and ensures more reliable ecological security assessments. Theoretically, this study advances methodological innovation in tourism ecological security evaluation by moving beyond static weighting and single-model designs. The proposed approach enables the dynamic tracking of fluctuations in the relative importance of ecosystem factors over time and under varying environmental conditions, thus filling a critical gap in the literature on the dynamic evolution of ecological security. Moreover, the incorporation of the grey cloud model allows the fuzziness and uncertainty of indicator data to be quantitatively characterized, overcoming the limitations of existing methods in handling small samples, incomplete information, and cognitive ambiguity, and thereby improving both the sensitivity and robustness of ecological security evaluations for island tourism destinations. This study recommends a zoning-based ecological governance strategy for the Zhoushan Archipelago, tailored to ecological security levels and carrying capacities to establish a gradient framework of "strict protection-moderate utilization-optimized enhancement." On heavily developed islands facing strong environmental pressures, visitor numbers should be controlled and tourism products redirected toward low-carbon and green transitions, while in less attractive areas ecological restoration should be coupled with the upgrading of eco-tourism and cultural experiences. A dynamic monitoring and tiered early-warning system, supported by remote sensing, drones, and IoT technologies, can provide real-time data on visitor flows, water quality, land use, and carbon emissions, enabling precise risk identification. The integration of the grey cloud model enhances risk detection under incomplete and uncertain data, and a cross-departmental platform ensures data sharing and dynamic updates. A smart management system further integrates monitoring, analysis, and decision support, offering real-time insights into tourist behavior and consumption trends, facilitating market regulation and visitor dispersion, and delivering timely alerts and governance recommendations.