PRIVACY-ACCURACY TRADE-OFF OF INFERENCE AS SERVICE

In this paper, we propose a general framework to provide a desirable trade-off between inference accuracy and privacy protection in the inference as service scenario. Instead of sending data directly to the server, the user will preprocess the data through a privacy-preserving mapping, which will increase privacy protection but reduce inference accuracy. To properly address the trade-off between privacy protection and inference accuracy, we formulate an optimization problem to find the optimal privacy-preserving mapping. Even though the problem is non-convex in general, we characterize nice structures of the problem and develop an iterative algorithm to find the desired privacy-preserving mapping.

Automated summary

This paper proposes a general framework to achieve a desirable trade-off between inference accuracy and privacy protection in the inference as service scenario, by preprocessing data through a privacy-preserving mapping to enhance privacy protection while reducing inference accuracy. An optimization problem is formulated to find the optimal privacy-preserving mapping, and an iterative algorithm is developed to obtain the desired privacy-preserving mapping by-characterizing nice structures of the problem.