A Novel Attack Scenario Dataset Collection for Intrusion Detection System in CAN Network.

As technology advances, the mechanical components of vehicles are being replaced by microcontrollers known as Electronic Control Units (ECUs). These ECUs connect to form an In-Vehicle Network, and they communicate control messages to each other in order to control the vehicle. There are several message protocols that can be used for IVN communication, but the most used protocol is the Controller Area Network (CAN) message. While CAN messages offer flexibility and cost-effectiveness to the network, their lack of security can make the network vulnerable to attacks, which can have serious consequences for both vehicles and drivers. In order to prevent such vulnerabilities, researchers have studied security systems such as Intrusion Detection Systems (IDS). However, the lack of availability and integrity of datasets can lead to misdetection, making it difficult to evaluate methods effectively. Therefore, we collected IVN messages based on several scenarios via collection tool. In this paper, we analyze CAN messages of each scenario and put into ML and DL models as testing dataset, and observe the results of each scenario. These findings shows miscalculation of the data collection scenario could lead to misdetection.