HaarAE: an unsupervised anomaly detection model for IOT devices based on Haar wavelet transform

Given the shortcomings of the existing anomaly detection methods based on IoT devices, including insufficient feature extraction, poor model fitting effect and low accuracy, this paper proposes an unsupervised IoT device traffic anomaly detection model called HaarAE, which introduces Haar wavelet transform to enhance the feature expression of original data and improve the model’s ability to identify anomalies. The convolutional autoencoder was used to construct the network structure, the memory module is introduced to increase the reconstruction error, and the ConvLSTM layer was added to the encoder to extract the temporal characteristics of the data. The output of each layer of decoder is cascaded with the output of the corresponding ConvLSTM layer, so that the decoder can obtain more coding information of each layer to reconstruct the original data and enhance the fitting ability of the model. Experiments on public datasets and real traffic datasets indicate that compared to the mainstream unsupervised models, HaarAE improves the anomaly detection effect.