A method and system for intelligent traffic scheduling protection

Abstract

The invention proposes an intelligent flow scheduling protection method and system. The method includes: based on the benign test data and the model structure of the traffic scheduling model to be detected, constructing a benign traffic scheduling model with a model structure and a traffic scheduling error rate lower than a first threshold, and building a model structure with a triggering data containing malicious attack The malicious traffic scheduling model of the trigger sample; input the trigger sample containing malicious attack data into the benign traffic scheduling model and the malicious traffic scheduling model to determine the index characteristics of the benign traffic scheduling model and the malicious traffic scheduling model; the benign The test data is input into the traffic scheduling model to be detected to obtain the characteristics of the indicators to be detected in the traffic scheduling model to be detected, and the traffic scheduling to be detected is determined based on the characteristics of the indicators to be detected, the characteristics of the indicators of the benign traffic scheduling model, and the characteristics of the indicators of the malignant traffic scheduling model The nature of the model to achieve intelligent traffic scheduling protection.

Description

Technical field [0001] The present invention belongs to the field of intelligent flow scheduling protection, in particular, to a smart flow scheduling protection method and system. Background technique [0002] In the flow schedule application scenario based on the circulating neural network, since the depth neural network is easily affected against disturbance and poisoning attacks, there is a problem with the problem of robby, which is easy to cause the traffic dispatch error. In particular, there is a safety problem such as interference and malicious attacks in the training phase and the inference stage of the circulating neural network, so that there is a safety problem such as traffic in the network environment, which will generate serious property losses in the intelligent network information system. [0003] In order to further improve the performance of traffic scheduling task in the intelligent network information system, it is necessary to make safety vulnerabilities in...

AI Technical Summary

Problems solved by technology

[0002] In the traffic scheduling application scenario based on the recurrent neural network, because the deep neural network is vulnerable to anti-perturbation and poisoning attacks, there is a problem of poor robustness of the model, which is likely to cause security problems in traffic scheduling errors

Especially in the training stage and reasoning stage of the cyclic neural network, there are anti-interference and malicious attacks, which cause security problems such as traffic surges in the network environment, and serious property losses will occur in the intelligent network information system.

Images