Research on Anomaly Detection Methods Based on Diffusion Models

TL;DR

This work develops an unsupervised anomaly detection framework based on diffusion probabilistic models that learns the distribution of normal data and reconstructs inputs through reverse diffusion. It integrates multi‑scale wavelet analysis, attention mechanisms, and a cross‑modal input strategy to handle image and audio data, using a reconstruction‑plus‑semantic discrepancy anomaly score. Empirical results on MVTec AD and time‑series benchmarks (NAB/UCR) show state‑of‑the‑art performance, with notable gains from the Wavelet Pyramid and long‑range temporal modeling, while remaining non‑adversarial in training. The approach offers a robust, modality‑agnostic solution with strong practical potential for real‑world inspection and monitoring tasks, while highlighting challenges in inference speed and interpretability that warrant future work.

Abstract

Anomaly detection is a fundamental task in machine learning and data mining, with significant applications in cybersecurity, industrial fault diagnosis, and clinical disease monitoring. Traditional methods, such as statistical modeling and machine learning-based approaches, often face challenges in handling complex, high-dimensional data distributions. In this study, we explore the potential of diffusion models for anomaly detection, proposing a novel framework that leverages the strengths of diffusion probabilistic models (DPMs) to effectively identify anomalies in both image and audio data. The proposed method models the distribution of normal data through a diffusion process and reconstructs input data via reverse diffusion, using a combination of reconstruction errors and semantic discrepancies as anomaly indicators. To enhance the framework's performance, we introduce multi-scale feature extraction, attention mechanisms, and wavelet-domain representations, enabling the model to capture fine-grained structures and global dependencies in the data. Extensive experiments on benchmark datasets, including MVTec AD and UrbanSound8K, demonstrate that our method outperforms state-of-the-art anomaly detection techniques, achieving superior accuracy and robustness across diverse data modalities. This research highlights the effectiveness of diffusion models in anomaly detection and provides a robust and efficient solution for real-world applications.