LoSplit: Loss-Guided Dynamic Split for Training-Time Defense Against Graph Backdoor Attacks

Di Jin, Yuxiang Zhang, Bingdao Feng, Xiaobao Wang, Dongxiao He, Zhen Wang

Keywords: GNNs, Training-Time Defense, Backdoor Defense

TL;DR: This paper proposes LoSplit, a training-time defense that detects and mitigates graph backdoors by dynamically analyzing early loss divergence.

Abstract:

Graph Neural Networks (GNNs) are vulnerable to backdoor attacks. Existing defenses primarily rely on detecting structural anomalies, distributional outliers, or perturbation-induced prediction instability, which struggle to handle the more subtle, feature-based attacks that do not introduce obvious topological changes. Our empirical analysis reveals that both structure-based and feature-based attacks not
only cause early loss convergence of target nodes but also induce a class-coherent loss drift, where this early convergence gradually spreads to nearby clean nodes, leading to significant distribution overlap. To address this issue, we propose LoSplit, the first training-time defense framework in graph that leverages this early-stage loss drift to accurately split target nodes. Our method dynamically selects epochs with maximal loss divergence, clusters target nodes via Gaussian Mixture Models (GMM), and applies a Decoupling-Forgetting strategy to break the association between target nodes and malicious label. Extensive experiments on multiple real-world datasets demonstrate the effectiveness of our approach, significantly reducing attack success rates while maintaining high clean accuracy across diverse backdoor attack strategies.

Primary Area: Deep learning (e.g., architectures, generative models, optimization for deep networks, foundation models, LLMs)

Submission Number: 11645