Multi-Layer Reconstruction Errors Autoencoding and Density Estimate for Network Anomaly Detection

Anomaly detection is an important method for intrusion detection.In recent years,unsupervised methods have been widely researched because they do not require labeling.For example,a nonlinear autoencoder can use reconstruction errors to attain the discrimination threshold.This method is not effective when the model complexity is high or the data contains noise.The method for detecting the density of compressed features in a hidden layer can be used to reduce the influence of noise on the selection of the threshold because the density of abnormal data in hidden layers is smaller than normal data.However,compressed features may lose some of the high-dimensional distribution information of the original data.In this paper,we present an efficient anomaly detection framework for unsupervised anomaly detection,which includes network data capturing,processing,feature extraction,and anomaly detection.We employ a deep autoencoder to obtain compressed features and multi-layer reconstruction errors,and feeds them the same to the Gaussian mixture model to estimate the density.The proposed approach is trained and tested on multiple current intrusion detection datasets and real network scenes,and performance indicators,namely accuracy,recall,and F1-score,are better than other autoencoder models.

Towards Human‑Vehicle Interaction: Driving Risk Analysis Under Different Driver Vigilance States and Driving Risk Detection Method

The driver's behavior plays a crucial role in transportation safety.It is widely acknowledged that driver vigilance is a major contributor to traffic accidents.However,the quantitative impact of driver vigilance on driving risk has yet to be fully explored.This study aims to investigate the relationship between driver vigilance and driving risk,using data recorded from 28 drivers who maintain a speed of 80 km/h on a monotonous highway for 2 hours.The k-means and linear fitting methods are used to analyze the driving risk distribution under different driver vigilance states.Additionally,this study proposes a research framework for analyzing driving risk and develops three classification models(KNN,SVM,and DNN)to recognize the driving risk status.The results show that the frequency of low-risk incidents is negatively correlated with the driver's vigilance level,whereas the frequency of moderate-risk and high-risk incidents is positively correlated with the driver's vigilance level.The DNN model performs the best,achieving an accuracy of 0.972,recall of 0.972,precision of 0.973,and f1-score of 0.972,compared to KNN and SVM.This research could serve as a valuable reference for the design of warning systems and intelligent vehicles.