Spatio-Temporal Attention Graph Neural Network: Explaining causalities with Attention
Kosti Koistinen, Kirsi Hellsten, Joni Herttuainen, Kimmo K. Kaski · 2026
Read Spatio-Temporal Attention Graph Neural Network: Explaining causalities with Attention by Kosti Koistinen, Kirsi Hellsten, Joni Herttuainen, Kimmo K. Kaski (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-10676.txt.
Introduction
Kosti Koistinen and co-authors propose an attention graph neural network, with space and time built in, for explainable anomaly detection in industrial control systems. Such systems face growing cyber-physical threats, while machine learning detectors suffer from poor explainability, false alarms, and drift. Their model treats sensors, controllers, and network entities as nodes in a learned graph, so attention can highlight the links behind an alert. It fuses SCADA, network flow, and payload data in one analysis. A conformal prediction layer controls false alarm rates and watches for drift.
Catalog record
| CREATOR | Kosti Koistinen, Kirsi Hellsten, Joni Herttuainen, Kimmo K. Kaski |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 121125 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.10676
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Spatio-Temporal Attention Graph Neural Network: Explaining causalities with Attention Kosti Koistinen, Kirsi Hellsten, Joni Herttuainen, Kimmo K. Kaski Abstract Industrial Control Systems (ICS) underpin critical infrastructure and face growing cyber–physical threats due to the convergence of operational technology and networked environments. While machine learning–based anomaly detection approaches in ICS shows strong theoretical performance, deployment is often limited by poor explainability, high false-positive rates, and sensitivity to evolving system behavior, i.e., baseline drifting. We propose a Spatio-Temporal Attention Graph Neural Network (STA-GNN) for unsupervised and explainable anomaly detection in ICS that models both temporal dynamics and relational structure of the system. Sensors, controllers, and network entities are represented as nodes in a dynamically learned graph, enabling the model to capture inter-dependencies across physical processes and communication patterns. Attention mechanisms provide influential relationships, supporting inspection of correlations and potential causal pathways behind detected events. The approach supports multiple data modalities, including SCADA point measurements, network flow features, and payload features, and thus enables unified cyber–physical analysis. To address operational requirements, we incorporate a conformal prediction strategy to control false alarm rates and monitor performance degradation under drifting of the environment. Our findings highlight the possibilities and limitations of model evaluation and common pitfalls in anomaly detection in ICS. Our findings emphasise the importance of explainable, drift-aware evaluation for reliable deployment of learning-based security monitoring systems. 1 Introduction Modern societies rely on uninterrupted functioning of interconnected critical infrastructure, such as electric power grids, water treatment plants, and manufacturing systems [ 15 ] . A disruption in these Operational Technology (OT) systems can cascade into severe economic, social, and physical consequences, from prolonged power outages to contaminated water supplies. Over the past decade, cyberattacks such as Stuxnet [ 30 ] , Industroyer [ 17 ] and the Colonial Pipeline incident [ 7 ] have demonstrated that threats once limited to Information Technology (IT) networks can now directly impact the physical world, such as equipment damage or even threats to human life [ 9 ] . During the past decade, cyberattacks on OT networks have been reported to have increased five fold from 300 annually to 1600 [ 12 ] . The actual scale is likely to be significantly higher, as many OT intrusions remain unreported or undiscovered due to limited monitoring capabilities. By 2024, operational disruption had become routine: 50–75% of ransomware incidents caused partial shutdowns, and approximately 25% resulted in complete production stoppages, causing significant financial damage [ 12 ] . Industrial Control Systems (ICS) form the technological backbone of critical infrastructure and are often the target of cyberattacks against OT systems. They regulate physical processes through sensors, actuators, and Programmable Logic Controllers (PLCs) and are maintained through Supervisory Control and Data Acquisition (SCADA). Traditionally isolated from external networks, the OT systems once enjoyed a degree of "security by separation." However, the shift toward networked automation, remote management, and the Industrial Internet of Things (IIoT) has converged IT and OT networks, allowing adversaries to move laterally from corporate systems to industrial environments. This development has exposed ICS environments to a wide spectrum of cyber threats [ 46 ] . In addition, OT environments often rely on legacy hardware, strict safety protocols, and systems that cannot be easily updated or patched, which further increases their vulnerability to cyberattacks. ICS threats range from malware infections and ransomware to unauthorised remote access, data manipulation, and process disruption. In many cases, attackers exploit vulnerabilities in outdated software, weak authentication, or insecure network configurations that were never designed with cybersecurity in mind. Typical weaknesses include outdated protocols that allow unintended access or manipulation of control traffic. The types of attacks are commonly divided into network-based and physical-based attacks [ 9 ] . The former includes Denial of Service (DoS), injection, and Man-in-the-Middle attacks, while the latter include stealth attacks, data tampering, and damage attacks. To detect and mitigate these complex attacks, Intrusion Detection Systems (IDS) are widely used in modern industrial cybersecurity. An IDS typically consists of the monitoring, pre-processing, and detection phases [ 53 ] . Among various detection approaches, such as signature-based, rule-based, and hybrid-based, anomaly-detection-based IDS have gained significant attention for their ability to learn normal operational behavior and discover deviations that may signal attacks, intrusions, or malfunctions [ 41 ] . In OT networks, this capability is crucial, as anomalies are often subtle irregularities rather than clear malicious signatures. They may appear as small fluctuations in sensor data, unexpected timing patterns, unusual command sequences, or deviations in process variables that remain within protocol limits but still indicate unsafe or suspicious behavior. There are various approaches that have been applied to detect and prevent cyber intrusions, including statistical modeling, Bayesian inference, and rule-based systems. These methods often rely on predefined assumptions about normal and abnormal behavior [ 23 ] . However, as industrial systems become more complex and dynamic, such fixed models struggle to capture the nonlinear and time-varying nature of real-world operations [ 11 ] . In contrast, machine learning–based approaches have attracted widespread interest for their ability to automatically learn patterns from data and adapt to evolving system behavior [ 41 ] . These methods can uncover correlations between multiple variables, making them particularly suitable for anomaly detection in ICS. Traditional machine learning approaches include, for example, k-nearest neighbors, Random Forests, and Support Vector Machines [ 4 ] . However, despite their efficiency in classification, these methods are insufficient to model temporal dependencies that are inherent in OT traffic. They are also sensitive to imbalanced datasets, such that a new unseen anomaly often remains undetected. Furthermore, in most OT environments, the majority of traffic is benign, while only a small fraction represents attacks. This imbalance can lead to biased classifiers that fail to detect rare but critical anomalies. To overcome these limitations, Deep learning approach has emerged as a promising solution. Autoregressive architectures such as Long Short-Term Memory (LSTM) networks [ 13 ] and other Recurrent Neural Network approaches (RNNs) [ 19 ] can capture complex temporal patterns among variables, which allows the model to understand how changes in one part of the system influence the rest. However, the performance of these models also suffers from unbalanced data. Some other methods include autoencoders, Generative Adversarial Networks, and a mixture of these models, but all have similar limitations. More recently, transformer-based architectures have become popular because of their ability to model long-range dependencies using self-attention mechanisms. Their success in natural language processing has motivated research into their application for anomaly detection in time-series and network data, where sequential dependencies and contextual relationships are crucial. Transformer models, and particularly adaptations of language models, show the potential to capture complex semantic patterns in network traffic representations [ 26 ] . Beyond and within sequential approaches, graph-based deep learning provides a fundamentally different way to represent and analyse OT systems. By modeling the system as a graph, where nodes represent entities (such as devices or sensors) and edges represent their relationships or communications, a more realistic and structured view of the environment can be obtained. Graph-based models are able to uncover non-linear correlations and long-range dependencies that traditional time-series or tabular approaches often miss. Graph Neural Networks (GNNs), such as Graph Convolutional Networks (GCNs) [ 27 ]