Spatio-Temporal Graphs Beyond Grids: Benchmark for Maritime Anomaly Detection

TL;DR

The paper tackles anomaly detection in non-grid spatio-temporal systems, focusing on maritime traffic where fixed spatial anchors are absent. It proposes a graph-based anomaly-detection benchmark by extending OMTAD to support node-, edge-, and graph-level anomalies, enabling systematic evaluation across granularities. A two-agent data-augmentation pipeline—Trajectory Synthesizer and Anomaly Injector—operates under a Coordinator to enrich inter-vessel context and inject semantically meaningful anomalies, using a shared perception schema. Preliminary experiments show GNN-based models outpace purely temporal baselines, underscoring the value of graph structure; the authors also plan to open-source the dataset and broaden anomaly definitions with future benchmarks across baselines.

Abstract

Spatio-temporal graph neural networks (ST-GNNs) have achieved notable success in structured domains such as road traffic and public transportation, where spatial entities can be naturally represented as fixed nodes. In contrast, many real-world systems including maritime traffic lack such fixed anchors, making the construction of spatio-temporal graphs a fundamental challenge. Anomaly detection in these non-grid environments is particularly difficult due to the absence of canonical reference points, the sparsity and irregularity of trajectories, and the fact that anomalies may manifest at multiple granularities. In this work, we introduce a novel benchmark dataset for anomaly detection in the maritime domain, extending the Open Maritime Traffic Analysis Dataset (OMTAD) into a benchmark tailored for graph-based anomaly detection. Our dataset enables systematic evaluation across three different granularities: node-level, edge-level, and graph-level anomalies. We plan to employ two specialized LLM-based agents: \emph{Trajectory Synthesizer} and \emph{Anomaly Injector} to construct richer interaction contexts and generate semantically meaningful anomalies. We expect this benchmark to promote reproducibility and to foster methodological advances in anomaly detection for non-grid spatio-temporal systems.

Figures (1)

• Figure 1: Preliminary results of time-series models and their GNN-integrated variants under different $r_{\text{traj}}$ settings. (a) LSTM-based models. (b) Transformer-based models. "TRANS" denotes Transformer.