Privacy-Preserving Travel Time Prediction for Internet of Vehicles: A Crowdsensing and Federated Learning Approach

Abstract

Travel time prediction (TTP) is an important module task to support various applications for Internet of Vehicles (IoVs). Although TTP has been widely investigated in the existing literature, most of them assume that the traffic data for estimating the travel time are comprehensive and public for free. However, accurate TTP needs real-time vehicular data so that the prediction can be adaptive to traffic changes. Moreover, since real-time data contain vehicles’ privacy, TTP requires protection during the data processing. In this paper, we propose a novel Privacy-Preserving TTP mechanism for IoVs, $\mathbb{P}\mathbb{T}$Prediction, based on crowdsensing and federated learning. In crowdsensing, a data curator continually collects traffic data from vehicles for TTP. To protect the vehicles’ privacy, we make use of the federated learning so that vehicles can help the data curator train the prediction model without revealing their information. We also design a spatial prefix encoding method to protect vehicles’ location information, along with a ciphertext-policy attribute-based encryption (CP-ABE) mechanism to protect the prediction model of the curator. We evaluate $\mathbb{P}\mathbb{T}$Prediction in terms of MAE, MSE, RMSE on two real-world traffic datasets. The experimental results illustrate that the proposed $\mathbb{P}\mathbb{T}$Prediction shows higher prediction accuracy and stronger privacy protection comparing to the existing methods.