Privacy-Preserving Language Model Inference with Instance Obfuscation

TL;DR

This paper introduces Instance-Obfuscated Inference (IOI), a novel method to protect decision privacy in language model inference, maintaining black-box access while ensuring data privacy with minimal overhead.

Contribution

The paper proposes a new privacy-preserving inference technique for language models that safeguards decision privacy without compromising model black-box functionality.

Findings

IOI effectively protects decision privacy in NLP tasks.

The method maintains high inference accuracy.

Low additional communication and computation overhead.

Abstract

Language Models as a Service (LMaaS) offers convenient access for developers and researchers to perform inference using pre-trained language models. Nonetheless, the input data and the inference results containing private information are exposed as plaintext during the service call, leading to privacy issues. Recent studies have started tackling the privacy issue by transforming input data into privacy-preserving representation from the user-end with the techniques such as noise addition and content perturbation, while the exploration of inference result protection, namely decision privacy, is still a blank page. In order to maintain the black-box manner of LMaaS, conducting data privacy protection, especially for the decision, is a challenging task because the process has to be seamless to the models and accompanied by limited communication and computation overhead. We thus propose Instance-Obfuscated Inference (IOI) method, which focuses on addressing the decision privacy issue of natural language understanding tasks in their complete life-cycle. Besides, we conduct comprehensive experiments to evaluate the performance as well as the privacy-protection strength of the proposed method on various benchmarking tasks.