Detection and Classification of Anomalies in Internet of Vehicles using Convolutional Neural Networks

Internet of Vehicles in brief is the distributed networks that makes the use of data produced due to connected cars and vehicular ad-hoc networks. With the progress in Science and Technology, The Internet of Vehicles has got an exponential boost, leading to tremendous progress in the field. The smart vehicles are equipped with various sensors which are data collection points, from which various useful data is collected for the processing and operating of the intelligent vehicles and play a vital role in implementation and communication of the vehicles. The smart vehicles has been a boon to humans which include fewer accidents, faster travel times, less theft due to tracking. But with these advantages, the vehicles have to cope up with various malicious intrusions and issues that compromise with the security and integrity of the vehicle. This is one of the major issues in the IOV which makes the data more vulnerable to be exposed. The research work aims to make the system more secure and safe by the detection and classification of the attacks that make these smart vehicles unsafe. The approach used in this work is based on the CNN based architecture VGG19, in which the data obtained from OBD port of vehicle and converted to visual format is feed. The types of the attacks included in this research work for the classification are Denial of Service Attack (DoS), Spoofing(RPM Spoof, Gear Spoof), Fuzzy Attack. The purpose of the research is to design a solution that makes the system more secure, preserves the privacy and increase the trustability of the user.