Diffuse to Detect: A Generalizable Framework for Anomaly Detection with Diffusion Models Applications to UAVs and Beyond

2025-10-26

AI Summary

To address low sensitivity, poor scalability, and insufficient dependency modeling in anomaly detection for high-dimensional spatiotemporal data (e.g., UAV sensor streams), this paper proposes *Diffuse to Detect*: the first framework to leverage a single-step diffusion model for anomaly detection, implicitly estimating the data score function via noise prediction. It integrates a graph neural network to dynamically capture spatiotemporal sensor dependencies and introduces a dual-branch architecture—comprising a parametric energy-scoring network and a nonparametric statistical test—to flexibly balance inference efficiency and interpretability. By circumventing the iterative sampling bottleneck of generative models, our method significantly reduces latency. Extensive experiments demonstrate state-of-the-art performance across UAV sensing, multivariate time-series, and image anomaly detection tasks, validating its cross-modal generalizability and industrial deployment readiness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Graph-based Machine Learning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Anomaly detection in complex, high-dimensional data, such as UAV sensor readings, is essential for operational safety but challenging for existing methods due to their limited sensitivity, scalability, and inability to capture intricate dependencies. We propose the Diffuse to Detect (DTD) framework, a novel approach that innovatively adapts diffusion models for anomaly detection, diverging from their conventional use in generative tasks with high inference time. By comparison, DTD employs a single-step diffusion process to predict noise patterns, enabling rapid and precise identification of anomalies without reconstruction errors. This approach is grounded in robust theoretical foundations that link noise prediction to the data distribution's score function, ensuring reliable deviation detection. By integrating Graph Neural Networks to model sensor relationships as dynamic graphs, DTD effectively captures spatial (inter-sensor) and temporal anomalies. Its two-branch architecture, with parametric neural network-based energy scoring for scalability and nonparametric statistical methods for interpretability, provides flexible trade-offs between computational efficiency and transparency. Extensive evaluations on UAV sensor data, multivariate time series, and images demonstrate DTD's superior performance over existing methods, underscoring its generality across diverse data modalities. This versatility, combined with its adaptability, positions DTD as a transformative solution for safety-critical applications, including industrial monitoring and beyond.

Problem

• Detecting anomalies in complex high-dimensional data like UAV sensor readings
• Overcoming limited sensitivity and scalability of existing anomaly detection methods
• Capturing intricate spatial and temporal dependencies in multivariate time series

Innovation

• Single-step diffusion process for rapid anomaly detection
• Graph Neural Networks modeling sensor relationships dynamically
• Two-branch architecture balancing scalability and interpretability