Privacy-Preserving Lane Change Prediction using Deep Learning Models

Lane Change Prediction (LCP) is a pivotal component of Intelligent Transportation Systems (ITS) which aims to enhance road safety and optimize traffic flow. Deep learning models have been used in LCP systems, but the need for extensive data poses privacy challenges in deploying such models. This paper aims to utilize Secure Multiparty Computation (SMPC) technique that allows multiple parties to jointly compute lane change probability using their private inputs without revealing sensitive information to each other. We trained Recurrent Neural Network (RNN) models on HighD dataset. In the inference part, Secure Tanh function was utilized for privacy-preserving LCP. Experimental results show that the accuracy of the proposed LCP based on SMPC is almost the same as the traditional LCP with unsecure computation, by calculating MSE for main function in non-SMPC and SMPC scenarios in the range of 10(-7).