Detecting systematic anomalies affecting systems when inputs are stationary time series

TL;DR

This paper introduces a novel anomaly detection method for control systems affected by systematic anomalies in stationary time series inputs, capable of handling diverse data types and providing interpretability through real-world and simulated experiments.

Contribution

The paper presents a new anomaly detection approach that works with a broad class of stationary time series and demonstrates its effectiveness through real and simulated data analyses.

Findings

Effective detection of anomalies in stationary time series

Applicable to diverse types of input data

Validated through real-world and controlled experiments

Abstract

We develop an anomaly-detection method when systematic anomalies, possibly statistically very similar to genuine inputs, are affecting control systems at the input and/or output stages. The method allows anomaly-free inputs (i.e., those before contamination) to originate from a wide class of random sequences, thus opening up possibilities for diverse applications. To illustrate how the method works on data, and how to interpret its results and make decisions, we analyze several actual time series, which are originally non-stationary but in the process of analysis are converted into stationary. As a further illustration, we provide a controlled experiment with anomaly-free inputs following an ARMA time series model under various contamination scenarios.