Research on privacy protection mechanism based on edge computing in Internet of Vehicles

Aiming at the contradiction between real-time data processing and privacy protection in edge computing of Internet of Vehicles, this paper proposes a dynamic collaborative differential privacy mechanism (DCDP). This mechanism clusters edge nodes by DBSCAN algorithm, builds a density-aware dynamic privacy budget allocation model, and realizes multi-node collaborative data aggregation by combining the improved PBFT consensus protocol. The experiment is based on the SUMO-NS-3 joint simulation platform. In the scenario of 300 m coverage radius, when the vehicle density reaches 80 vehicles $/ \mathrm{km}^{2}$, the privacy leakage rate of DCDP is only $0.078 \%$, which is $42.3 \%$ lower than that of traditional differential privacy (TDP); the root mean square error (RMSE) of traffic flow estimation is 2.15 vehicles/minute, which is $28.7 \%$ higher than that of the location-independent algorithm (PIDP). The system delay experiment shows that the average processing time of DCDP is 12.4 ms, which is $65.2 \%$ shorter than that of the blockchain-assisted solution (BCDP).