<p>Data sharing across organizations is fundamental to data-driven analytics and decision-making; however, shared datasets often contain sensitive information and are vulnerable to privacy leakage and unauthorized redistribution. Ensuring both privacy protection and ownership verification remains a significant challenge, as most existing approaches address these concerns independently. This paper proposes Selective Attribute Local Differential Privacy with Dynamic Watermarking (SA-LDP-DW), a unified framework for privacy-aware data publishing that preserves analytical utility. The proposed framework applies local differential privacy by dynamically allocating attribute-level privacy budgets based on data sensitivity, enabling controlled perturbation while maintaining the effectiveness of downstream analytical tasks. To support data governance and ownership verification, a dynamic watermarking mechanism is embedded into the data publishing process, ensuring data integrity under common data modification scenarios. The framework is evaluated on four publicly available datasets, namely the UCI Adult, Credit Card Default, Pima Diabetes, and California Housing datasets, using both classification and regression tasks. Experimental results demonstrate a mean absolute error of 0.120 and an R² of approximately 0.84, indicating strong analytical performance. The watermarking mechanism achieves a detection accuracy greater than 0.94 with a bit error rate below 0.04, even under record-deletion and value-scaling attacks. Compared with existing methods, the proposed approach incurs less than 2.4% utility degradation. These results demonstrate that SA-LDP-DW supports responsible data sharing, data governance, and privacy-aware data analytics, enabling privacy protection, ownership verification, and reliable analytical utility for real-world data-driven applications.</p>