Parallel acceleration method for KNN algorithm-based network intrusion detection

Abstract

The invention provides a parallel acceleration method for KNN algorithm-based network intrusion detection. A CUDA parallel computing model is adopted in the method. The method comprises the steps thatfirstly, parallel analysis is conducted on KNN algorithm-based network intrusion detection, and when the distances between network intrusion detection data points and a training data set are computed, a general matrix multiplication function provided by the CUDA is adopted for acceleration, so that the computing speed is increased; secondly, in the distance sorting stage, a selection mechanism for two parallel sorting strategies is provided, the sorting algorithm needing the shorter sorting time can be flexibly selected for distance sorting according to the sorting result of a few data; and finally, in the classification stage of the intrusion detection data points, counting is conducted by adopting a CUDA-based atom addition operation. The experiment result shows that the acceleration method is valid, and the parallel acceleration performance of network intrusion detection is effectively improved on the condition that the detection rate is guaranteed.

Description

(1) Technical field [0001] The invention relates to a network intrusion detection technology in the field of information security, and is a parallelized acceleration method for network intrusion detection based on a KNN algorithm. (2) Background technology [0002] The purpose of network intrusion detection is to analyze the data traffic transmitted on the network, and find and detect abnormal traffic from it. At present, there are many classification algorithms used in network intrusion detection systems. Among them, network intrusion detection based on KNN (K-Nearest Neighbor) classification algorithm is the most commonly used. The most classic solution method of the KNN algorithm is the brute force search method. This method first calculates the distance between each network intrusion detection data point and the network intrusion detection training data set in turn, and then quickly sorts the distance of each intrusion detection data point in turn, thereby Get the k training...

AI Technical Summary

Problems solved by technology

[0006] The present invention first aims at the complex distance calculation problem in the KNN classification algorithm, extracts the common part of the calculation, and uses the general matrix multiplication function provided by CUDA to realize the distance calculation between the network intrusion detection data point and the network intrusion detection training data set , which improves the operation speed

Images