Satellite altimetry provides an important data source for recovering the marine gravity field. However, the conventional remove-compute-restore (RCR) strategy commonly used for deriving marine deflection of the vertical (DOV) causes the long-wavelength components of the derived DOV to depend to some extent on reference gravity field models. The two-dimensional (2D) sea surface height observations from SWOT/KaRIn wide-swath altimetry provide a new opportunity for constructing marine DOV datasets without relying on the conventional RCR strategy. Here, we construct a global 1′ × 1′ marine DOV dataset, named SDUST2025DOV_SWOT, based on multi-cycle SWOT/KaRIn wide-swath altimetry observations and a local geoid fitting approach. Based on SDUST2025DOV_SWOT, the corresponding marine gravity anomaly dataset SDUST2025GRA_SWOT is further derived. The datasets cover global ocean regions between 77.6°S and 77.6°N. The quality of the datasets is evaluated through comparisons with existing global products, including SIO V33.1DOV/GRA, SWOT_04DOV/GRA, and DTU21GRA, as well as independent shipborne gravity observations. The results show that SDUST2025DOV_SWOT agrees well with existing DOV products in both components. Correlation analyses between DOV datasets and seafloor topographic slopes in typical regions with different seafloor topographic characteristics indicate that SDUST2025DOV_SWOT and SWOT_04DOV exhibit stronger consistency with short-wavelength seafloor topographic signals than SIO V33.1DOV. Independent shipborne gravity observations show that the RMS difference between SDUST2025GRA_SWOT and shipborne gravity anomalies is 3.82 mGal at the global scale, which is reduced by 0.71, 0.22, and 0.31 mGal compared with SIO V33.1GRA, DTU21GRA, and SWOT_04GRA, respectively. Regional assessments further demonstrate that the dataset achieves more pronounced improvements in areas with complex seafloor topography. In a region containing both seamounts and trenches, the RMS difference of SDUST2025GRA_SWOT is reduced by up to 0.90 mGal compared with SIO V33.1GRA. The SDUST2025DOV_SWOT and SDUST2025GRA_SWOT are available at <a href="https://doi.org/10.5281/zenodo.21408557">https://doi.org/10.5281/zenodo.21408557</a><span> </span>(Zhang et al., 2026).