Trajectory Privacy Protection Based On Deep Learning

In today's society,Location-Based Service(Location-Based Service,LBS)is developing rapidly and is widely used in APPs of various smart mobile terminals.LBS providers and customers use trajectory data as the information entity and the Internet as the information carrier for information interaction.During the entire process of information interaction,a large amount of trajectory data with rich spatiotemporal information is generated.On the one hand,LBS brings great aspects to people's clothing,food,housing,and transportation.On the other hand,the issue of trajectory privacy leakage accompanying LBS will bring huge losses to users' economic security and even personal security.Therefore,how to effectively protect the privacy of the user's trajectory is of great significance to the further development of LBS.Because there is no need for a trusted third-party server,and the accurate service of the LBS provider and the small computational cost are the advantages,the dummy trajectory method is widely used to protect the user's trajectory privacy.Researchers have proposed many trajectory privacy protection schemes based on the dummy trajectory method,but the existing dummy trajectory method only considers the geometric meaning of the trajectory,and does not consider the hidden human movement model in the trajectory,that is,the generated dummy trajectory and the real trajectory are not distributed in the same time and space.Our experiment proves that using the dummy trajectory discriminator based on neural network can easily crack the existing dummy trajectory generation scheme.If the adversary masters this technology,the privacy of the user's trajectory will be easily exposed.The existing dummy trajectory scheme cannot effectively protect the privacy of the user's trajectory.This paper proposes a dummy trajectory generation algorithm based on neural network.The main work of this article includes:(1)Propose a dummy trajectory recognition scheme based on deep learning,and construct some dummy trajectory recognizers based on neural network models.Then put the real and dummy trajectories into the recognizer for training to identify the dummy trajectories generated by the traditional dummy trajectory generation algorithm,to prove that the traditional dummy trajectory generation algorithm cannot resist the recognition attack of the dummy trajectory discriminator based on the neural network.(2)Aiming at the main problems of the current personal privacy protection method of dummy trajectory,a dummy trajectory generation scheme based on neural network is clearly proposed.Based on the traditional algorithm for generating dummy trajectory,the scheme adds the steps of resisting map recognition and resisting discriminator recognition based on deep learning,so that the generated dummy trajectories are both spatially distributed with the real trajectories,and will not fall into unreachable area.(3)Through experimental simulation,it is concluded that the dummy trajectory discriminator based on the neural network can identify more than 96% of the dummy trajectories generated by the existing dummy trajectory generation schemes,which proves that the existing dummy trajectory generation schemes cannot protect user's trajectory privacy well.Then the dummy trajectory generation scheme proposed in this paper is used to generate dummy trajectories,and the recognition rate of the generated dummy trajectories is only 16.5%.Thus,the user's trajectory privacy is well protected.