HW-GNN: Homophily-Aware Gaussian-Window Constrained Graph Spectral Network for Social Network Bot Detection

TL;DR

HW-GNN tackles bot detection by addressing limitations of broad-spectrum spectral GNNs through a Gaussian-window spectral network that concentrates on bot-discriminative frequency bands. It injects domain knowledge via a homophily-aware adaptation mechanism, using an MLP-driven parameter learning pipeline and a frequency distribution loss to align spectral focus with network structure. The method demonstrates state-of-the-art F1-score improvements (average 4.3%) across five benchmarks and remains compatible as a plug-in for existing spectral GNNs. This approach enhances robustness to heterophily and scalability by leveraging polynomial approximations and a multi-band filtering strategy, offering practical impact for scalable, accurate social bot detection. Future work includes dynamic networks and potential integration with large language models for improved detection capabilities.

Abstract

Social bots are increasingly polluting online platforms by spreading misinformation and engaging in coordinated manipulation, posing severe threats to cybersecurity. Graph Neural Networks (GNNs) have become mainstream for social bot detection due to their ability to integrate structural and attribute features, with spectral-based approaches demonstrating particular efficacy due to discriminative patterns in the spectral domain. However, current spectral GNN methods face two limitations: (1) their broad-spectrum fitting mechanisms degrade the focus on bot-specific spectral features, and (2) certain domain knowledge valuable for bot detection, e.g., low homophily correlates with high-frequency features, has not been fully incorporated into existing methods. To address these challenges, we propose HW-GNN, a novel homophily-aware graph spectral network with Gaussian window constraints. Our framework introduces two key innovations: (i) a Gaussian-window constrained spectral network that employs learnable Gaussian windows to highlight bot-related spectral features, and (ii) a homophily-aware adaptation mechanism that injects domain knowledge between homophily ratios and frequency features into the Gaussian window optimization process. Through extensive experimentation on multiple benchmark datasets, we demonstrate that HW-GNN achieves state-of-the-art bot detection performance, outperforming existing methods with an average improvement of 4.3% in F1-score, while exhibiting strong plug-in compatibility with existing spectral GNNs.

Figures (4)

• Figure 1: Illustration of homophily and spectral graph filtering in social bot detection. Homophilic community $A$: nodes with same labels form dense connections and concentrate spectral energy in low frequencies. Heterophilic community $B$: cross-label connections create dispersed patterns, shifting spectral energy toward higher frequencies. The spectral representations are then processed using polynomial basis functions to construct graph filters for bot detection.
• Figure 2: HW-GNN Framework: homophily-aware Gaussian-window constrained graph spectral network for social bot detection. The framework employs learnable Gaussian windows to modulate the weights of polynomial basis functions, enabling focused spectral analysis on bot-discriminative frequency bands. The homophily-aware adaptation mechanism injects domain knowledge between homophily ratios and frequency features to guide window parameter learning.
• Figure 3: Sensitivity analysis of key parameters in HW-GNN. (a)--(c): F1-score on Twibot20, Twibot22, and MGTAB datasets, respectively, with varying polynomial order $K$ and different numbers of Gaussian windows $S$ (each line represents a different $S$). (d): F1-score variation with respect to the frequency distribution loss weight $\lambda_f$.
• Figure 4: Learned spectral filter $h(\lambda)$ for different datasets by HW-GNN (blue) and original Bernstein polynomial (black).