Beyond Surface-Level Patterns: An Essence-Driven Defense Framework Against Jailbreak Attacks in LLMs

TL;DR

This paper introduces EDDF, a novel essence-driven defense framework that enhances LLMs' robustness against jailbreak attacks by focusing on attack core patterns rather than surface features.

Contribution

The paper proposes a plug-and-play, two-stage defense method that extracts and stores attack essences for improved detection of adversarial prompts in LLMs.

Findings

EDDF reduces attack success rate by at least 20%.

It outperforms existing defenses significantly.

The framework effectively captures attack core patterns.

Abstract

Although Aligned Large Language Models (LLMs) are trained to refuse harmful requests, they remain vulnerable to jailbreak attacks. Unfortunately, existing methods often focus on surface-level patterns, overlooking the deeper attack essences. As a result, defenses fail when attack prompts change, even though the underlying "attack essence" remains the same. To address this issue, we introduce EDDF, an \textbf{E}ssence-\textbf{D}riven \textbf{D}efense \textbf{F}ramework Against Jailbreak Attacks in LLMs. EDDF is a plug-and-play input-filtering method and operates in two stages: 1) offline essence database construction, and 2) online adversarial query detection. The key idea behind EDDF is to extract the "attack essence" from a diverse set of known attack instances and store it in an offline vector database. Experimental results demonstrate that EDDF significantly outperforms existing methods by reducing the Attack Success Rate by at least 20\%, underscoring its superior robustness against jailbreak attacks.