In the recent years, many studies claim that humans have a unique driving behavior style that could be used as a fingerprint in recognizing the identity of the driver. With the rising evolution of Machine Learning (ML), the research efforts aiming to take advantage of the human driving style identifiers have been increasing exponentially. For Advanced Driver Assistance Systems (ADAS), this attribute can be an efficient factor to ensure the security and protection of the vehicle. Additionally, it extends the ADAS capabilities by creating different profiles for the drivers, which helps every driver according to his own driving style and improve the ADAS fidelity. Nonetheless, certain problems in the unpredictability of human behavior and the effectiveness of capturing the temporal features of the signal represented an ongoing challenge to accomplish driver identification. In this paper, we propose a novel deep learning approach to driver identification based on a Residual Convolutional Network (RCN). This approach outperforms the existing state of the art methods in less than two hours of training, while simultaneously achieving 99.3% accuracy. The used data are exclusively provided by the Controller Area Network (CAN-Bus) vehicle data that eliminates any privacy invading concerns from the user.

近年来，许多研究声称人类具有独特的驾驶行为方式，可以用作识别驾驶员身份的指纹。随着机器学习（ML）的不断发展，旨在利用人类驾驶风格标识符的研究工作呈指数级增长。对于高级驾驶员辅助系统（ADAS），此属性可以是确保车辆安全和保护的有效因素。此外，它通过为驾驶员创建不同的配置文件来扩展ADAS功能，从而帮助每个驾驶员根据自己的驾驶风格并提高ADAS的保真度。尽管如此，人类行为不可预测性中的某些问题以及捕获信号的时间特征的有效性代表了完成驾驶员识别的一项持续挑战。在本文中，我们提出了一种基于残差卷积网络（RCN）的驾驶员识别的新型深度学习方法。这种方法在不到两个小时的训练中就优于现有的现有方法，同时达到了99.3％的准确性。所使用的数据仅由控制器局域网（CAN-Bus）车辆数据提供，从而消除了用户侵犯隐私的任何担忧。这种方法在不到两个小时的训练中就优于现有的现有方法，同时达到了99.3％的准确性。所使用的数据仅由控制器局域网（CAN-Bus）车辆数据提供，从而消除了用户侵犯隐私的任何担忧。这种方法在不到两个小时的训练中就优于现有的现有方法，同时达到了99.3％的准确性。所使用的数据仅由控制器局域网（CAN-Bus）车辆数据提供，从而消除了用户侵犯隐私的任何担忧。