Prompt Leakage effect and defense strategies for multi-turn LLM interactions

TL;DR

This paper systematically investigates prompt leakage vulnerabilities in multi-turn LLM interactions, evaluates attack success rates, and proposes defense strategies to enhance security and privacy in LLM applications.

Contribution

It introduces a novel threat model leveraging LLM sycophancy, evaluates leakage across multiple models and domains, and assesses various defense strategies including finetuning and black-box methods.

Findings

Attack success rate increased from 17.7% to 86.2% under the threat model.

Different defenses can mitigate prompt leakage with varying costs.

Analysis provides insights for building secure multi-turn LLM systems.

Abstract

Prompt leakage poses a compelling security and privacy threat in LLM applications. Leakage of system prompts may compromise intellectual property, and act as adversarial reconnaissance for an attacker. A systematic evaluation of prompt leakage threats and mitigation strategies is lacking, especially for multi-turn LLM interactions. In this paper, we systematically investigate LLM vulnerabilities against prompt leakage for 10 closed- and open-source LLMs, across four domains. We design a unique threat model which leverages the LLM sycophancy effect and elevates the average attack success rate (ASR) from 17.7% to 86.2% in a multi-turn setting. Our standardized setup further allows dissecting leakage of specific prompt contents such as task instructions and knowledge documents. We measure the mitigation effect of 7 black-box defense strategies, along with finetuning an open-source model to defend against leakage attempts. We present different combination of defenses against our threat model, including a cost analysis. Our study highlights key takeaways for building secure LLM applications and provides directions for research in multi-turn LLM interactions