Perseus: Leveraging Common Data Patterns with Curriculum Learning for More Robust Graph Neural Networks

TL;DR

Perseus is a novel adversarial defense method based on curriculum learning that assesses edge difficulty using global homophily and applies a curriculum learning strategy to adjust the learning order, guiding the model to learn the full graph structure while adaptively focusing on common data patterns.

Abstract

Graph Neural Networks (GNNs) excel at handling graph data but remain vulnerable to adversarial attacks. Existing defense methods typically rely on assumptions like graph sparsity and homophily to either preprocess the graph or guide structure learning. However, preprocessing methods often struggle to accurately distinguish between normal edges and adversarial perturbations, leading to suboptimal results due to the loss of valuable edge information. Robust graph neural network models train directly on graph data affected by adversarial perturbations, without preprocessing. This can cause the model to get stuck in poor local optima, negatively affecting its performance. To address these challenges, we propose Perseus, a novel adversarial defense method based on curriculum learning. Perseus assesses edge difficulty using global homophily and applies a curriculum learning strategy to adjust the learning order, guiding the model to learn the full graph structure while adaptively focusing on common data patterns. This approach mitigates the impact of adversarial perturbations. Experiments show that models trained with Perseus achieve superior performance and are significantly more robust to adversarial attacks.

Figures (4)

• Figure 1: Performance of different metrics on Cora under Metattack with 20% perturbation rates
• Figure 2: The overall pipeline of Perseus. (a) Quantifying the edge difficulty based on the original image, the edge selection module takes topk edges based on the current residual graph. Furthermore, the residual graph is obtained from the edge quantization index and the current selected edge. The timing of performing edge selection is subsequently determined by judging the model convergence state. (b) Iterative learning process of Perseus. The model is trained starting from the initial simple structure and gradually adds more multiple optimal edges until the training structure converges to the input structure.
• Figure 3: Performance of different models under Metattack with different perturbation rates
• Figure 4: Train loss and validation loss of Cora under Metattack with 0.25 perturbation.