Abstract: With the growth of ship navigation data sharing and intelligent shipping application demand, ships face challenges such as privacy leakage and data utility degradation in trajectory publishing and data services. Higher requirements are imposed on the security and data utility of navigation data protection mechanisms. Existing ship trajectory privacy protection methods still have deficiencies in dynamic feature representation and prediction error suppression. It is difficult to balance privacy protection, data utility, and arrival time prediction accuracy. To address these problems, a ship navigation data protection and publishing mechanism based on differential privacy was proposed. First, feature extraction was performed on ship navigation states based on the three-degree-of-freedom motion model of the ship’s horizontal-plane motion. Then, a dynamic privacy budget allocation strategy was constructed. The privacy budget was adaptively adjusted according to dynamic features such as changes in ship speed and heading. A deep neural network was also employed to compensate for speed perturbation, so as to reduce the influence of privacy noise on data prediction performance. The proposed method was verified on a real ship trajectory dataset. The results show that, compared with unperturbed trajectories, the increase in arrival time prediction error of this method can be controlled within 0.5 h. High data utility is maintained while ship trajectory privacy is effectively protected. A feasible technical path is provided for secure data publishing in intelligent shipping.