Differential Privacy Trajectory Data Publishing Method Based on RNN

User trajectory data is critical for developing location services applications and urban traffic management. However, user trajectory data contains sensitive information such as home and workplace, and the direct release of user trajectory data will raise serious privacy concerns. To solve this problem, differential privacy is an effective method to synthesize trajectory data. However, the existing trajectory publishing model only considers the relationship between the front and back trajectory points, and lacks flexibility in trajectory data extraction. Therefore, this paper proposes a new privacy publishing method for trajectory data, which combines the parameter minimum description algorithm with Dijkstra's method to select the characteristic trajectory points representing the entire trajectory. This method further improves the flexibility of trajectory selection, and the relationship between any two trajectory points is more closely linked by recurrent neural network. Experimental results show that our solution improves the availability of published trajectory datasets while protecting user privacy.