TRAFFIC ANOMALY DETECTION IN VEHICLE BUS BY RECURRENT LSTM NEURAL NETWORK

Modern high-end cars have many electronic control units for driving assistance that combine huge amounts of data about the functioning of car components. A significant part of these vehicles use a controller area network for communication between electronic units. Controller area network is a simple and reliable network protocol that due to its simplicity lacks any security mechanisms for data transmission. The problem of controller area network vulnerability is worsening as constantly growing amounts of data between cars, road infrastructure and the Internet. The traffic of attacks on controller area networks can be treated as abnormal that allows using anomaly detection methods for their recognition. In this work we propose the recurrent long short-term memory encoder-decoder neural network for controller area network attacks detection.