An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks

TL;DR

This paper introduces an interpretable N-gram perplexity-based threat model to evaluate and compare jailbreak attacks on large language models, revealing that many attacks are less effective than previously thought and often exploit rare bigrams.

Contribution

The paper presents a novel, LLM-agnostic, nonparametric threat model based on N-gram perplexity for benchmarking jailbreak attacks on LLMs, enabling transparent and comprehensive analysis.

Findings

Attack success rates are lower than previously reported.

Discrete optimization attacks outperform LLM-based attacks.

Effective attacks exploit infrequent or dataset-specific bigrams.

Abstract

A plethora of jailbreaking attacks have been proposed to obtain harmful responses from safety-tuned LLMs. These methods largely succeed in coercing the target output in their original settings, but their attacks vary substantially in fluency and computational effort. In this work, we propose a unified threat model for the principled comparison of these methods. Our threat model checks if a given jailbreak is likely to occur in the distribution of text. For this, we build an N-gram language model on 1T tokens, which, unlike model-based perplexity, allows for an LLM-agnostic, nonparametric, and inherently interpretable evaluation. We adapt popular attacks to this threat model, and, for the first time, benchmark these attacks on equal footing with it. After an extensive comparison, we find attack success rates against safety-tuned modern models to be lower than previously presented and that attacks based on discrete optimization significantly outperform recent LLM-based attacks. Being inherently interpretable, our threat model allows for a comprehensive analysis and comparison of jailbreak attacks. We find that effective attacks exploit and abuse infrequent bigrams, either selecting the ones absent from real-world text or rare ones, e.g., specific to Reddit or code datasets.