Abstract
Against the backdrop of the rapid development of the Internet of Things and vehicle technology, the emergence of intelligent connected vehicles provides a technological foundation for vehicle crowdsensing (VCS). This emerging sensing paradigm is beneficial for reducing financial and time costs related to data collection and improving data quality. Inappropriate user selection may lead to redundant sensing data, consequently degrading the quality of sensing tasks. At present, user selection schemes have the following lacks: (1) Privacy protection and data security are often neglected, resulting in reduced user participation willingness. (2) Insufficient data quality assurance and challenges in processing high-dimensional redundant data remain significant issues. (3) Key exchange protocols rely on traditional cryptographic algorithms, which struggle to comply with Chinese cryptographic standards and incur high overhead. In order to solve these problems, this article proposes a secure and efficient user selection scheme in vehicular crowdsensing, named SEUS-VCS, a novel framework that pioneers secure and efficient user selection in VCS. It adopts a key exchange protocol based on Shang Mi Two (SM2) digital signature to ensure the security of data transmission, including the correctness of signatures and session keys. To protect user privacy, pseudonyms are generated to achieve user anonymity while ensuring traceability. In addition, a prediction model based on principal component analysis-enhanced long short-term memory (PCA-Enhanced LSTM) model is constructed by combining the advantages of principal component analysis (PCA) and long short-term memory (LSTM) network. The model uses PCA for data dimensionality reduction to eliminate redundant information and then employs LSTM to process time series data, capture long-term dependencies for more accurate user credit prediction, screen high-quality users, and improve perceived data quality. Security analysis demonstrates that under the Canetti-Krawczyk (CK) model and the Decisional Diffie-Hellman (DDH) security assumption, the SEUS-VCS scheme satisfies the security of session keys. Performance evaluation indicates that compared to other key exchange schemes, the SEUS-VCS scheme shows a significant advantage in supporting national cryptographic algorithms and anonymity. Compared to the scheme with the lowest computational cost and the scheme with the lowest communication cost, the SEUS-VCS scheme reduces computational cost by 42% and decreases communication cost by 50%. The SEUS-VCS scheme has advantages in reducing loss function (Loss), Mean Square Error (MSE), and Mean Absolute Error (MAE), and the predicted results match the true data very well. Performance evaluation indicates that compared to the LSTM baseline, which has the best MSE and MAE performance, the MSE and MAE of the SEUS-VCS scheme are reduced by 47% and 10%, respectively. This scheme is not only applicable to VCS, but can also provide reference for other fields involving credit prediction and incentive mechanisms.