Train communication network flow identification method based on deep learning

Abstract

The invention discloses a train communication network flow identification method based on deep learning, and the method specifically comprises the steps: collecting train communication network flow data through a package capturing tool, reading and analyzing a pcap file, carrying out the session recombination, data cleaning, equal-length processing and normalization processing, carrying out the annotation of a label for sample data, and training the one-dimensional convolutional neural network by using the sample data; and finally, achieving the effective classification of the traditional application layer protocols and the railway private protocols in the train communication network traffic only by inputting unknown traffic into the trained model. According to the invention, effective identification and classification of common application layer protocols and railway private protocols in a train control and service network are realized; useful feature information can be extracted more accurately, and higher precision, precision ratio and recall ratio are achieved.

Description

technical field [0001] The invention belongs to the technical field of information communication, and in particular relates to a deep learning-based traffic identification method of a train communication network. Background technique [0002] With the integration of traditional train communication network and passenger service network and the introduction of Ethernet technology, the requirements for transmission performance of train communication network have been greatly increased. In addition, many security problems have been caused, such as Dos attack and IP address spoofing. In order to ensure the reasonable scheduling and supervision of traffic in the train service network and avoid network congestion, network intrusion and other phenomena, it is necessary to classify and identify network traffic and analyze its characteristics. The current traffic identification methods are mainly divided into the following directions: identification method based on port number, identi...

AI Technical Summary

Problems solved by technology

However, this type of method is not without disadvantages. For example, some algorithms have high complexity and high computational overhead, and the scalability, robustness, and real-time performance of the classifier must also be considered.

Images