A trajectory protection method based on differential privacy and semantic attributes

Abstract

With the popularity of smart wearable devices, Location-Based Services (LBS) have been widely applied. However, LBS generates thousands of trajectories, which may potentially leak personal information. To address such privacy concerns, it's often necessary to protect users' trajectory data. Existing differential privacy schemes commonly used for trajectory protection suffer from inefficiency and lack support for semantic attributes. To tackle these issues, this paper proposes a trajectory protection method based on differential privacy and semantic attributes (STrajGAN). Firstly, we employ differential privacy to process the raw data, enhancing efficiency while preserving privacy. Subsequently, the processed data is introduced into a Generative Adversarial Network (GAN) with Gated Recurrent Units (GRU), which generates similar trajectory data while considering the influence of semantic attributes of trajectory points, thus safeguarding user trajectory privacy. Experimental results demonstrate that, compared to other models, STrajGAN can provide better trajectory privacy protection by considering the influence of trajectory point semantic attributes.