Feature selection algorithm of network attack big data under the interference of fading noise

In order to detect the existence of big data with aggressive behavior in the network, it is necessary to study the feature selection algorithm of network attack big data. When the current feature selection algorithm is used to select the big data features with aggressive behavior in the network, the interference is highly impacted by fading noise, and the calculated feature importance is inconsistent with reality. But these algorithms have poor anti-interference performance and low accuracy of feature selection. This paper proposes a feature selection algorithm of network attack big data under the interference of fading noise, removes the fading noise in the network through wavelet threshold denoising method, and uses the self-coding deep neural network method to reduce the dimensionality of network attack big data. Based on the results of data dimensionality reduction, the mutual information matrix is calculated and normalized to obtain a random matrix with the same form as the mutual information matrix. The importance of big data features is obtained by the mutual information matrix and the characteristic correlation matrix of the random matrix. According to the importance of features, the sorting of data features is completed, and features of high importance are selected as characteristics of network attack big data to complete feature selection of network attack big data under the interference of fading noise. Experimental results show that the proposed method has strong anti-interference and high feature selection accuracy. © 2019 Informa UK Limited, trading as Taylor & Francis Group.