A system for confidential training and inference for vertically partitioned dataset using secure multi-party computation /
Title
A system for confidential training and inference for vertically partitioned dataset using secure multi-party computation /
Subject
Computer Science
Description
Patent Number: 202241057386, Application: Kapil Tiwari.
The present invention applies to probabilistic inference employing secure multi-party computation (SMPC) methodologies including confidential training, inference, and model protection for vertically partitioned datasets. The proposed invention addresses privacy complexities of vertically partitioned data owners, model owners, and ML clients using confidential training, inference, and model protection (CoTraIn-VPD), a technique to boost data, model parameter, and client input privacy during ML inference without affecting accuracy or performance.
The present invention applies to probabilistic inference employing secure multi-party computation (SMPC) methodologies including confidential training, inference, and model protection for vertically partitioned datasets. The proposed invention addresses privacy complexities of vertically partitioned data owners, model owners, and ML clients using confidential training, inference, and model protection (CoTraIn-VPD), a technique to boost data, model parameter, and client input privacy during ML inference without affecting accuracy or performance.
Creator
Shukla, Samiksha.
Publisher
Intellectual Property India
Date
2022
Language
English
Type
Patent
Collection
Citation
Shukla, Samiksha., “A system for confidential training and inference for vertically partitioned dataset using secure multi-party computation /,” CHRIST (Deemed To Be University) Institutional Repository, accessed December 22, 2024, https://archives.christuniversity.in/items/show/2793.