VCC vehicle attack detection method based on deep learning

Abstract

The invention discloses a VCC vehicle attack detection method based on deep learning. The method comprises internal abnormal vehicle detection and external abnormal vehicle detection. The internal abnormal vehicle detection comprises the following steps: continuously collecting information of VCC internal vehicles and carrying out preprocessing; training an auto-encoder by using the preprocessed data; using the trained model to detect the abnormity of the internal vehicle; the external abnormal vehicle detection comprises the following steps: preprocessing information of an external vehicle requesting to join a VCC; extracting VCC internal vehicle information closest to the external vehicle application time as normal vehicle data; extracting the features of the external vehicle information and the internal normal vehicle information by using the coding part of the trained auto-encoder; constructing a support vector data description classifier and performing training; and detecting an external vehicle by using the trained support vector data description classifier. According to the method, non-supervision VCC abnormal vehicle detection is realized by using the spatial-temporal characteristics of the vehicle information.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a VCC vehicle attack detection method based on deep learning. Background technique [0002] Vehicular Cloud Computing (VCC), VCC integrates the calculation of idle vehicles. Communication and storage capabilities, combined with VANETs communication constitute a mobile cloud computing. Among them, vehicles are both resource providers and resource users, and multiple vehicles cooperate to perform tasks. However, unlike traditional cloud computing, VCC is composed of multiple vehicles to provide cloud resources. Normal vehicles and attacking vehicles have the same access rights. They can share various resources provided by the cloud computing platform, so it is difficult to deploy security protection strategies. At the same time, the rapid topology changes and open communication network of vehicles make the data transmitted in VCC more vulnerable to attacks. [0003] With the r...

AI Technical Summary

Problems solved by technology

However, statistics-based methods have poor detection performance on high-dimensional data, and rule-based methods can achieve better classification results, but the process of detection and decision-making is extremely complicated, and the detection results often depend on the business decision-making level and ability of experts. And it requires a lot of manpower. The establishment of detection models based on machine learning methods is currently the most mainstream research direction. However, with the expansion of network capacity and more and more unlabeled data, the advantages of traditional machine learning methods are no longer available. exist

Images