PriDyG: Privacy-preserving Dynamic Graph Inference with LLM-GNN Collaboration

2026-08-04

AI Summary

This work addresses the challenge of cumulative edge-level privacy leakage in dynamic graph reasoning caused by frequent updates. To tackle this issue, the authors propose an incremental private multi-hop aggregation mechanism that integrates GNN-based structural learning with LLM-enhanced semantic reasoning, enabling efficient handling of graph evolution while preserving edge-level differential privacy. The key innovation lies in achieving a total privacy cost independent of the number of updates—a first in this domain—and leveraging large language models to enrich node semantics without incurring additional edge-level privacy overhead. Experimental results demonstrate that the proposed method significantly outperforms geometric-decay baselines across four benchmark tasks, attaining utility close to non-private retraining approaches while reducing cumulative privacy cost by up to three orders of magnitude.

Abstract

Graph inference over relational data can expose sensitive edge information, and this risk becomes more severe in dynamic graphs, where repeated model updates cause privacy loss to accumulate. We formulate Edge-level Differentially Private Dynamic Graph Inference (EDG) and propose PriDyG, a private inference framework that combines GNN-based structural learning with LLM-based semantic reasoning. PriDyG introduces incremental private multi-hop aggregation, which buffers newly arrived edges and processes each edge exactly once. By parallel composition, the total privacy cost equals that of a single static release, independent of the number or schedule of model updates. Compared with geometrically decaying budget allocation, incremental aggregation avoids exponentially increasing noise while preserving exact one-hop signals and at least half of two-hop information transfers. PriDyG further complements privatized GNN outputs with LLM predictions derived solely from node text, incurring no additional edge-level privacy cost. Experiments on four benchmarks for node classification and link prediction show that PriDyG consistently outperforms geometrically decaying baselines under the same privacy budget and matches the utility of naive per-update retraining while reducing cumulative privacy cost by up to three orders of magnitude.

Problem

• dynamic graph
• privacy preservation
• edge-level privacy
• differential privacy
• graph inference

Innovation

• Edge-level Differential Privacy
• Dynamic Graph Inference
• Incremental Private Aggregation
• GNN-LLM Collaboration
• Parallel Composition