The paper proposes OnePath, a secure and efficient framework for cloud-based decision tree inference. The key highlights are:
Secure Encryption of Decision Trees: The framework ensures the encryption of decision trees while preserving their functionality for cloud-based inference. It creates node indexes and uses pseudo-random functions to conceal feature indexes, maintaining the decision tree's integrity and privacy.
Secure Internal Node Evaluation: The protocol transforms the comparison function at each internal node into a linear equation and applies functional encryption for its secure computation. This streamlines the evaluation while protecting the data.
Selective Traversal of Prediction Path: The framework obscures the true prediction path by evaluating only several internal nodes linearly related to the tree's depth. This significantly boosts computational efficiency and strengthens privacy.
Offline User Support: The protocol allows both providers and clients to perform secure inference without the need to remain online continuously, a critical advantage for real-world applications.
The authors provide formal security proofs and experimental results demonstrating the efficiency and practicality of the proposed framework, making it a promising solution for secure cloud-based decision tree inference.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Shuai Yuan, ... at arxiv.org 10-01-2024
https://arxiv.org/pdf/2409.19334.pdfDeeper Inquiries