Anomaly Detection and Generation with Diffusion Models: A Survey

2025-06-11

AI Summary

Anomaly detection faces fundamental challenges including scarcity of authentic anomaly samples and difficulty in modeling complex data distributions. To address these, this paper introduces the first systematic “Anomaly Detection and Generation协同” (ADGDM) unified framework, uncovering the intrinsic bidirectional enhancement mechanisms between detection and generation—namely, score matching, conditional sampling, and architectural coupling. We propose a fine-grained taxonomy that integrates diffusion model theory, unsupervised anomaly scoring functions, multimodal alignment techniques, and vision-language/ large language model (VLM/LLM) integration, supporting diverse modalities including images, videos, time-series, tabular, and multimodal data. This constitutes the most comprehensive ADGDM methodology to date. We explicitly identify three critical bottlenecks: scalability, computational efficiency, and generalization capability. Finally, we outline key future research directions, including efficient architecture design, dynamic conditional modeling, and foundation model collaboration.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Anomaly detection (AD) plays a pivotal role across diverse domains, including cybersecurity, finance, healthcare, and industrial manufacturing, by identifying unexpected patterns that deviate from established norms in real-world data. Recent advancements in deep learning, specifically diffusion models (DMs), have sparked significant interest due to their ability to learn complex data distributions and generate high-fidelity samples, offering a robust framework for unsupervised AD. In this survey, we comprehensively review anomaly detection and generation with diffusion models (ADGDM), presenting a tutorial-style analysis of the theoretical foundations and practical implementations and spanning images, videos, time series, tabular, and multimodal data. Crucially, unlike existing surveys that often treat anomaly detection and generation as separate problems, we highlight their inherent synergistic relationship. We reveal how DMs enable a reinforcing cycle where generation techniques directly address the fundamental challenge of anomaly data scarcity, while detection methods provide critical feedback to improve generation fidelity and relevance, advancing both capabilities beyond their individual potential. A detailed taxonomy categorizes ADGDM methods based on anomaly scoring mechanisms, conditioning strategies, and architectural designs, analyzing their strengths and limitations. We final discuss key challenges including scalability and computational efficiency, and outline promising future directions such as efficient architectures, conditioning strategies, and integration with foundation models (e.g., visual-language models and large language models). By synthesizing recent advances and outlining open research questions, this survey aims to guide researchers and practitioners in leveraging DMs for innovative AD solutions across diverse applications.

Problem

• Surveying anomaly detection and generation using diffusion models across domains
• Exploring synergy between anomaly detection and generation via diffusion models
• Addressing challenges like scalability in diffusion-based anomaly detection methods

Innovation

• Uses diffusion models for anomaly detection
• Combines anomaly detection and generation synergistically
• Addresses anomaly data scarcity via generation