Federated Time-Frequency Fusion Attention Network (Fed-TFANet) for Privacy-Preserving and Explainable Anomaly Detection in ICS

View at publisher Summarise with VIRAT
Authors: Sivakumar Depuru, Shaik Mulla Sony, Shaik Muhammad Faheem, Peddakotla Jaya Dev, Samuru Vishnu
Cited by: 0
References: 13

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Intelligent monitoring is becoming a major part of the Industrial Control Systems (ICS) to identify the presence of operational anomalies due to cyber-physical attacks. However, the centralized learning models are vulnerable to privacy issues and lack interpretability. In this research, the propose an improved architecture, Fed-TFANet, which is a follow-up of Time Frequency Fusion Feature Attention Network (TFANet) that incorporates federated learning and explainable attention encoding. Fed-TFANet allows distributed training of a model on distributed industrial nodes without data transfer, which does not expose local privacy and still allows analyzing global anomalies. Moreover, a built-in explainability module renders attribution maps in a visually engaging format that shows the patterns of sensors contributing to them. The three benchmark datasets, including SWaT, WADI, and OTS, are experimentally proven to show that Fed-TFANet has a 7.3 -fold increase in detection correctness, a 9.1-fold increase in recall, and a reduction in communication latency by 18.5% over the baseline TFANet. The interpretability index can be explained with SHAP scores 21% higher than with standard transformerbased approaches. The findings verify that the suggested FedTFANet has been able to realize secure, accurate, and interpretable anomaly detection that can be applied to nextgeneration intelligent industrial infrastructures.

Fields of Study & Subjects

Fault Detection and Control Systems Anomaly Detection Techniques and Applications Network Security and Intrusion Detection

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
2166–2173
Publication date
21 January 2026
Language
EN