An Intrusion Detection Method For Internet Of Vehicles Based On Fogcloud Collaborative Computing

Following the rapid development of 5G and its application in the Internet of Vehicles（Io V）,issues pertaining to its security and privacy are taken seriously.Connected cars and driverless cars are emerging industries,and safety measures and regulations are still being improved.In August 2020,Baidu’s driverless vehicles carried out the second round of road tests in Beijing,which means that driverless technology is one step closer to the practical application.With the deployment of Internet of Vehicles and driverless driving,once the vehicle is attacked,it will not only cause the risk of data leakage,but also threaten the life safety of the driver.Moreover,traditional cloud computing services cannot meet the requirements of low delay and high bandwidth of Internet of Vehicles.Therefore,thesis proposes an Io V intrusion detection system based on fog-cloud collaborative computing.Firstly,an intrusion detection architecture based on cloud and fog collaborative computing is proposed.The computing tasks are assigned according to the computing and storage capacities of different devices.Secondly,in this architecture,decision tree,convolutional neural network,genetic algorithm and other technologies are used to detect attacks.The main work contents of this paper are as follows:First,owing to the different computing and storage capabilities of fog nodes and cloud servers,a fog-cloud cooperative defense architecture that a)divides traffic data into normal and suspicious fog-nodal data,and b)classifies suspicious data on cloud servers to identify attack types was proposed.Second,aiming at the problem of the resource-limited fog nodes and changeable network environment,an improved pruning algorithm considering decision tree size and error rate is proposed to reduce the resource occupancy of fog nodes.Finally,to cope with the imbalance data problem,a GA-Cost-Sensitive Computational Neural Network model to classify the suspicious data was proposed.Genetic algorithm is used to automatically search for the optimal cost matrix.Specific classification of suspicious data can reduce the missing report of a few attacks.Simulations were conducted based on a real-world vehicular dataset,and their outcomes indicate that the proposed intrusion detection system scheme can achieve high performance with low-resource requirements.