Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Panther: A Cost-Effective Privacy-Preserving Framework for GNN Training and Inference Services in Cloud Environments

Congcong Chen, Xinyu Liu, Kaifeng Huang, Lifei Wei, Yang Shi

TL;DR

Panther introduces a four-party asynchronous MPC framework with random neighbor padding to enable cost-effective privacy-preserving training and inference for GNNs in cloud environments. It combines data preprocessing to obfuscate graph structure with secure building blocks (Hadamard, MatMult, AA, DReLU, ReLU, Softmax, Div, InvSqrt) to protect features, adjacency, weights, and outputs. Empirical results on Cora, CiteSeer, and PubMed show Panther achieves comparable accuracy to plaintext and outperforms state-of-the-art SecGNN in both time and communication, while reducing cloud costs by roughly half to two-fifths depending on the plan. Panther also demonstrates favorable scalability for large graphs and presents deployment considerations, including multi-cloud and federated strategies, making privacy-preserving GNN services more practical for real-world cloud use.

Abstract

Graph Neural Networks (GNNs) have marked significant impact in traffic state prediction, social recommendation, knowledge-aware question answering and so on. As more and more users move towards cloud computing, it has become a critical issue to unleash the power of GNNs while protecting the privacy in cloud environments. Specifically, the training data and inference data for GNNs need to be protected from being stolen by external adversaries. Meanwhile, the financial cost of cloud computing is another primary concern for users. Therefore, although existing studies have proposed privacy-preserving techniques for GNNs in cloud environments, their additional computational and communication overhead remain relatively high, causing high financial costs that limit their widespread adoption among users. To protect GNN privacy while lowering the additional financial costs, we introduce Panther, a cost-effective privacy-preserving framework for GNN training and inference services in cloud environments. Technically, Panther leverages four-party computation to asynchronously executing the secure array access protocol, and randomly pads the neighbor information of GNN nodes. We prove that Panther can protect privacy for both training and inference of GNN models. Our evaluation shows that Panther reduces the training and inference time by an average of 75.28% and 82.80%, respectively, and communication overhead by an average of 52.61% and 50.26% compared with the state-of-the-art, which is estimated to save an average of 55.05% and 59.00% in financial costs (based on on-demand pricing model) for the GNN training and inference process on Google Cloud Platform.

Panther: A Cost-Effective Privacy-Preserving Framework for GNN Training and Inference Services in Cloud Environments

TL;DR

Panther introduces a four-party asynchronous MPC framework with random neighbor padding to enable cost-effective privacy-preserving training and inference for GNNs in cloud environments. It combines data preprocessing to obfuscate graph structure with secure building blocks (Hadamard, MatMult, AA, DReLU, ReLU, Softmax, Div, InvSqrt) to protect features, adjacency, weights, and outputs. Empirical results on Cora, CiteSeer, and PubMed show Panther achieves comparable accuracy to plaintext and outperforms state-of-the-art SecGNN in both time and communication, while reducing cloud costs by roughly half to two-fifths depending on the plan. Panther also demonstrates favorable scalability for large graphs and presents deployment considerations, including multi-cloud and federated strategies, making privacy-preserving GNN services more practical for real-world cloud use.

Abstract

Graph Neural Networks (GNNs) have marked significant impact in traffic state prediction, social recommendation, knowledge-aware question answering and so on. As more and more users move towards cloud computing, it has become a critical issue to unleash the power of GNNs while protecting the privacy in cloud environments. Specifically, the training data and inference data for GNNs need to be protected from being stolen by external adversaries. Meanwhile, the financial cost of cloud computing is another primary concern for users. Therefore, although existing studies have proposed privacy-preserving techniques for GNNs in cloud environments, their additional computational and communication overhead remain relatively high, causing high financial costs that limit their widespread adoption among users. To protect GNN privacy while lowering the additional financial costs, we introduce Panther, a cost-effective privacy-preserving framework for GNN training and inference services in cloud environments. Technically, Panther leverages four-party computation to asynchronously executing the secure array access protocol, and randomly pads the neighbor information of GNN nodes. We prove that Panther can protect privacy for both training and inference of GNN models. Our evaluation shows that Panther reduces the training and inference time by an average of 75.28% and 82.80%, respectively, and communication overhead by an average of 52.61% and 50.26% compared with the state-of-the-art, which is estimated to save an average of 55.05% and 59.00% in financial costs (based on on-demand pricing model) for the GNN training and inference process on Google Cloud Platform.

Paper Structure

This paper contains 58 sections, 7 theorems, 9 equations, 14 figures, 14 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Concerns in Cloud Computing
  4. Graph Neural Networks
  5. Secret Sharing
  6. Panther
  7. Overview
  8. System Design
  9. Data preprocessing
  10. Multi-party computation
  11. Threat Model
  12. Security Analysis
  13. Protocol Constructions
  14. Basic Operations
  15. Building Blocks
  16. ...and 43 more sections

Key Result

Theorem 1

Protocol $\Pi_{\mathsf{Mult}}$ securely realizes $\mathcal{F}_{\mathsf{Mult}}$ (see Fig. 8, Appendix B) in the presence of one semi-honest corrupted party.

Figures (14)

  • Figure 1: GNN training/inference process for Panther (using the secure array access protocol as an example in the multi-party computation phase).
  • Figure 2: Similarities and differences between the secure neighbor information padding methods of Panther and SecGNN.
  • Figure 3: Secure array access protocol for Panther.
  • Figure 4: Efficiency comparison of SecGNN and Panther training phase.
  • Figure 5: Efficiency comparison of SecGNN and Panther inference phase.
  • ...and 9 more figures

Theorems & Definitions (8)

  • Theorem 1
  • Theorem 2
  • Theorem 3
  • Theorem 4
  • Theorem 5
  • Theorem 6
  • Theorem 7
  • Definition 1