Paper 2023/1807
Entrada to Secure Graph Convolutional Networks
Abstract
Graph convolutional networks (GCNs) are gaining popularity due to their powerful modelling capabilities. However, guaranteeing privacy is an issue when evaluating on inputs that contain users’ sensitive information such as financial transactions, medical records, etc. To address such privacy concerns, we design Entrada, a framework for securely evaluating GCNs that relies on the technique of secure multiparty computation (MPC). For efficiency and accuracy reasons, Entrada builds over the MPC framework of Tetrad (NDSS’22) and enhances the same by providing the necessary primitives. Moreover, Entrada leverages the GraphSC paradigm of Araki et al. (CCS’21) to further enhance efficiency. This entails designing a secure and efficient shuffle protocol specifically in the 4-party setting, which to the best of our knowledge, is done for the first time and may be of independent interest. Through extensive experiments, we showcase that the accuracy of secure GCN evaluated via Entrada is on par with its cleartext counterpart. We also benchmark efficiency of Entrada with respect to the included primitives as well as the framework as a whole. Finally, we showcase Entrada’s practicality by benchmarking GCN-based fraud detection application.
Metadata
- Available format(s)
- Category
- Cryptographic protocols
- Publication info
- Preprint.
- Keywords
- secure multiparty computationgraph convolutional networkssecure shuffle
- Contact author(s)
-
kotis @ iisc ac in
varshak @ iisc ac in
arpita @ iisc ac in
bhavishraj @ iisc ac in - History
- 2023-11-24: approved
- 2023-11-23: received
- See all versions
- Short URL
- https://ia.cr/2023/1807
- License
-
CC BY
BibTeX
@misc{cryptoeprint:2023/1807, author = {Nishat Koti and Varsha Bhat Kukkala and Arpita Patra and Bhavish Raj Gopal}, title = {Entrada to Secure Graph Convolutional Networks}, howpublished = {Cryptology {ePrint} Archive, Paper 2023/1807}, year = {2023}, url = {https://eprint.iacr.org/2023/1807} }