Ruslan Abdulin | Searcharxiv

arXiv2026

A Transformer-Based Mixture-of-Experts Framework for False Data Injection Attack Detection and Localization

False data injection attacks (FDIAs) threaten smart-grid operation by manipulating measurement data and misleading power system state estimation. Although recent data-driven methods have shown promising performance, many existing approaches rely on a single graph-filtering mechanism and therefore struggle to adapt to diverse cyberattack patterns. This paper proposes a topology-aware encoder-only transformer mixture-of-experts framework for joint FDIA detection and localization. The proposed model integrates Laplacian positional encoding and diffusion-biased self-attention to capture global spatial dependencies, while a densely routed mixture-of-experts module composed of Chebyshev and autoregressive moving-average graph convolutional experts adaptively applies complementary graph filters based on the current grid state. The framework is evaluated on the IEEE 118- and 300-bus systems using the New York Independent System Operator load profile. Experimental results demonstrate that the proposed approach achieves up to 93.91% detection F1 score and 84.97% localization F1 score, outperforming existing benchmark methods while maintaining low false-alarm rates.

math.OC↗

arXiv2026

Attention-Enhanced Graph Filtering for False Data Injection Attack Detection and Localization

The increasing deployment of Internet-of-Things (IoT)-enabled measurement devices in modern power systems has expanded the cyberattack surface of the grid. As a result, this critical infrastructure is increasingly exposed to cyberattacks, including false data injection attacks (FDIAs) that compromise measurement integrity and threaten reliable system operation. Existing FDIA detection methods primarily exploit spatial correlations and network topology using graph-based learning; however, these approaches often rely on high-dimensional representations and shallow classifiers, limiting their ability to capture local structural dependencies and global contextual relationships. Moreover, naively incorporating Transformer architectures can result in overly deep models that struggle to model localized grid dynamics. This paper proposes a joint FDIA detection and localization framework that integrates auto-regressive moving average (ARMA) graph convolutional filters with an Encoder-Only Transformer architecture. The ARMA-based graph filters provide robust, topology-aware feature extraction and adaptability to abrupt spectral changes, while the Transformer encoder leverages self-attention to capture long-range dependencies among grid elements without sacrificing essential local context. The proposed method is evaluated using real-world load data from the New York Independent System Operator (NYISO) applied to the IEEE 14- and 300-bus systems. Numerical results demonstrate that the proposed model effectively exploits both the state and topology of the power grid, achieving high accuracy in detecting FDIA events and localizing compromised nodes.

cs.LG↗