Deep Autoencoders with Value-at-Risk Thresholding for Unsupervised Anomaly Detection

TL;DR

This paper introduces a deep autoencoder with a novel Value-at-Risk thresholding method for real-time unsupervised anomaly detection, demonstrating its effectiveness against existing subspace methods on change detection datasets.

Contribution

It proposes a new VaR-based thresholding mechanism integrated with deep autoencoders for streaming anomaly detection, advancing the state-of-the-art in incremental learning approaches.

Findings

The proposed method outperforms traditional subspace methods on change detection datasets.

VaR thresholding effectively identifies anomalies in streaming data.

Deep autoencoders with VaR provide robust unsupervised anomaly detection.

Abstract

Many real-world monitoring and surveillance applications require non-trivial anomaly detection to be run in the streaming model. We consider an incremental-learning approach, wherein a deep-autoencoding (DAE) model of what is normal is trained and used to detect anomalies at the same time. In the detection of anomalies, we utilise a novel thresholding mechanism, based on value at risk (VaR). We compare the resulting convolutional neural network (CNN) against a number of subspace methods, and present results on changedetection net.