Feature-Based Trajectory Privacy Preserving via Low-Rank and Sparse Decomposition

The disclosure of sensitive contents hidden in trajectories may jeopardize individuals' privacy security. The privacy preserving technologies on trajectories face the challenge of how to give consideration to the spatiotemporal structure of mobile data. Traditional trajectory privacy preserving tricks focus chiefly on partial structure, while ignoring global feature of trajectories. This research orientation tends to cause exorbitant distortion on the spatiotemporal structure of trajectories, which may give rise to low data utility. The moving behavior has itself innate sparseness feature. Thus a trajectory privacy preserving method based on feature maintaining is proposed by introducing the low-rank and sparse decomposition technique about large matrix. The proposed method iteratively decompose the feature matrix until the rank achieving stability to extract the primary feature and eliminate the private parts. The released trajectories are reconstructed by perturbing the final refined feature matrix with a series of low-rank components generated in the decomposition procedure. Experimental results on real-world dataset verified that the proposed method has low information loss on large-scale data.