The controller area network (CAN) is the de facto standard for in-vehicle networks. Numerous vehicles are equipped with a CAN for the networking of electronic control units. Since the CAN has no intrinsic secured identification protocol, however, it is vulnerable to security attacks. In this paper, we propose a convolution neural network (CNN)-based message source identifier. Since the proposed scheme uses the physical characteristics of the CAN bus channel, it can be implemented without modification of the CAN protocol.