AR-Pro: Counterfactual Explanations for Anomaly Repair with Formal Properties
It addresses the problem of making anomaly detection more interpretable for users in domains like vision and time-series, offering a domain-independent framework, though it builds on existing methods and generative models.
The paper tackles the lack of interpretability in anomaly detection by introducing counterfactual explanations that generate non-anomalous versions of inputs using diffusion-based repairs, demonstrating effectiveness on vision and time-series datasets.
Anomaly detection is widely used for identifying critical errors and suspicious behaviors, but current methods lack interpretability. We leverage common properties of existing methods and recent advances in generative models to introduce counterfactual explanations for anomaly detection. Given an input, we generate its counterfactual as a diffusion-based repair that shows what a non-anomalous version should have looked like. A key advantage of this approach is that it enables a domain-independent formal specification of explainability desiderata, offering a unified framework for generating and evaluating explanations. We demonstrate the effectiveness of our anomaly explainability framework, AR-Pro, on vision (MVTec, VisA) and time-series (SWaT, WADI, HAI) anomaly datasets. The code used for the experiments is accessible at: https://github.com/xjiae/arpro.