An Evidence-Based Post-Hoc Adjustment Framework for Anomaly Detection Under Data Contamination

Sukanya Patra, Souhaib Ben Taieb

Keywords: anomaly detection, anomaly, contamination, test-time adaptation

Abstract: Unsupervised anomaly detection (AD) methods typically assume clean training data, yet real-world datasets often contain undetected or mislabeled anomalies, leading to significant performance degradation. Existing solutions require access to the training data, model pipeline or model parameters, limiting real-world applicability. To address this challenge, we propose EPHAD, a simple yet effective test-time adaptation framework that updates the outputs of AD models trained on contaminated datasets using evidence gathered at inference. Our approach integrates the prior captured by the AD model trained on the contaminated dataset with the output of an auxiliary evidence function at test-time using exponential tilting. This evidence can be derived from foundation models like CLIP, classical methods such as the Latent Outlier Factor or domain-specific knowledge. We validate its effectiveness through extensive experiments across eight image-based AD datasets, twenty-seven tabular datasets, and a real-world industrial dataset. Our code is publicly available https://github.com/sukanyapatra1997/EPAF.

Submission Number: 37