Lightweight Consensus Mechanisms and Graph Neural Networks for Decentralized Detection of Stealth Data Manipulation in High-Density Advanced Metering Infrastructures

• Adaan Ahsun

  Covenant University

  Author

Advanced Metering Infrastructure (AMI) deployments have expanded rapidly, introducing critical vulnerabilities to stealthy false data injection (FDI) attacks that evade conventional detection mechanisms. This study addresses the dual challenge of achieving high detection accuracy for subtle data manipulations while maintaining computational feasibility in resource-constrained, high-density metering environments. We propose a decentralized detection framework integrating lightweight consensus mechanisms with graph neural networks (GNNs) for topology-aware anomaly detection. The methodology combines a reputation-based lightweight consensus protocol for tamper-resistant data aggregation with a Graph Convolutional Network that models spatial dependencies among smart meters. Experiments on synthetic attack injections into smart meter consumption data demonstrate that the proposed framework achieves 89.4% detection accuracy for stealth attacks with low-magnitude manipulations, outperforming centralized baseline approaches by 12.7 percentage points while reducing communication overhead by 43%. The framework achieves 91.2% precision and 87.6% recall, with consensus layer verification adding negligible latency (mean 47ms per aggregation round). The results establish that decentralized, topology-aware detection architectures can effectively identify stealthy FDI attacks without imposing prohibitive computational burdens on AMI infrastructure. This work provides a replicable framework for utility operators seeking to enhance grid cybersecurity while respecting the resource constraints of high-density metering deployments.

• pdf

Copyright (c) 2026 Adaan Ahsun (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.