Differential Privacy Protection Method based on Deep Learning Model

Inhalt:
In order to protect the security of spatio-temporal trajectory data distribution, we propose a scheme based on deep learning model and differential privacy protection technology. Firstly, we divide the trajectory data into two-dimensional grid regions, count the trajectory density values of the grid, and design a top-down region merging method to obtain the division structure rank matrix. The features of spatio-temporal trajectory data are extracted by a deep learning model to achieve the prediction of the division structure rank matrix, and then the privacy protection of the trajectory data is achieved by distributing the privacy budget through differential privacy protection and adding Laplace noise by random sampling. From the experiments carried out it can be concluded that under the premise of satisfying differential privacy, the relative error of the proposed method is smaller, and the upper and lower limit values are more stable, which concludes that the propoesd method is better than the traditional methods UG and AG, and proves that this paper's method can be applied to a large amount of spatio-temporal trajectory data.