Anomaly detection method based on LSTM-Autoencoder and double feature extraction method
SUN Xuri1
LIU Mingfeng1
CHENG Hui1
PENG Bo1
ZHAO Yufei2
Abstract：In order to solve the problem of low detection accuracy and high false positive rate in a-nomaly detection. A Long Short Term Memory(LSTM) based anomaly detection method is proposed. First，the features are extracted from data packages and the session flows. To enrich the data features，the Discrete Wavelet Transform(DWT) technique is used to decompose the original data into the feature vectors with higher dimension. Considering the non-human abnormal data in the real network environment，the Grubbs Criterion is used to eliminate the non-human abnormal data in case of disturbing modeling LSTM-Autoencoder. Then，the reconstruction errors of input feature vectors are calculated by the LSTM-Autoencoder model. The distribution of the reconstruction errors is fitted and the detection threshold is determined. At last，the experiments are conducted on the real network data. The influences of the model structure and the environment noise on the detection performance are analyzed. The experimental results verified the feasibility of the proposed method and shown that the proposed method can effectively identify the abnormal data and has a better performance in detection accuracy compared with other detection methods.
Keywords：information safetylong short term memorydiscrete wavelet transformautoencoderGrubbs criteriondata reconstructionanomaly detection
Publication Date：2020-04-28
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：10( 17-26 )