Toward Scalable Graph Unlearning: A Node Influence Maximization based Approach

TL;DR

This work addresses graph unlearning in graph neural networks by tackling gradient-driven entanglement and scalability on web-scale graphs. It bridges graph unlearning with social influence maximization by introducing Node Influence Maximization (NIM), a decoupled influence-propagation mechanism with a fine-grained influence function to identify high-influence entities. Building on this, Scalable Graph Unlearning (SGU) offers lightweight, entity-specific fine-tuning with prototype representations and contrastive learning to balance forgetting and reasoning. Extensive experiments across 14 datasets, including billion-scale graphs, demonstrate competitive forgetting, robust inference protection, and strong scalability, establishing SGU as a practical and effective GU framework.

Abstract

Machine unlearning, as a pivotal technology for enhancing model robustness and data privacy, has garnered significant attention in prevalent web mining applications, especially in thriving graph-based scenarios. However, most existing graph unlearning (GU) approaches face significant challenges due to the intricate interactions among web-scale graph elements during the model training: (1) The gradient-driven node entanglement hinders the complete knowledge removal in response to unlearning requests; (2) The billion-level graph elements in the web scenarios present inevitable scalability issues. To break the above limitations, we open up a new perspective by drawing a connection between GU and conventional social influence maximization. To this end, we propose Node Influence Maximization (NIM) through the decoupled influence propagation model and fine-grained influence function in a scalable manner, which is crafted to be a plug-and-play strategy to identify potential nodes affected by unlearning entities. This approach enables offline execution independent of GU, allowing it to be seamlessly integrated into most GU methods to improve their unlearning performance. Based on this, we introduce Scalable Graph Unlearning (SGU) as a new fine-tuned framework, which balances the forgetting and reasoning capability of the unlearned model by entity-specific optimizations. Extensive experiments on 14 datasets, including large-scale ogbn-papers100M, have demonstrated the effectiveness of our approach. Specifically, NIM enhances the forgetting capability of most GU methods, while SGU achieves comprehensive SOTA performance and maintains scalability.

Figures (9)

• Figure 1: (a,b) Performance in the ogbn-arxiv (169k Nodes). The red line (AUC=0.5) represents the best unlearning performance under Membership Inference Attack shokri2017mia_attack. (c) Potential limitations of HIE selection under the learning-based GU.
• Figure 2: Overview of our proposed SGU framework.
• Figure 3: Node-level performance on arxiv within Edge Attack. The x-axis is the ratio of noisy edges to the existing edges.
• Figure 4: Predictive performance of node unlearning within different ratios on PPI.
• Figure 5: Sensitive analysis within GAMLP and node removal.

Theorems & Definitions (3)

• Definition 1
• Definition 2
• Definition 3