AutoAdv: Automated Adversarial Prompting for Multi-Turn Jailbreaking of Large Language Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Reddy, Aashray, Zagula, Andrew, Saban, Nicholas
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911331049799680
author Reddy, Aashray
Zagula, Andrew
Saban, Nicholas
author_facet Reddy, Aashray
Zagula, Andrew
Saban, Nicholas
contents Large Language Models (LLMs) remain vulnerable to jailbreaking attacks where adversarial prompts elicit harmful outputs. Yet most evaluations focus on single-turn interactions while real-world attacks unfold through adaptive multi-turn conversations. We present AutoAdv, a training-free framework for automated multi-turn jailbreaking that achieves an attack success rate of up to 95% on Llama-3.1-8B within six turns, a 24% improvement over single-turn baselines. AutoAdv uniquely combines three adaptive mechanisms: a pattern manager that learns from successful attacks to enhance future prompts, a temperature manager that dynamically adjusts sampling parameters based on failure modes, and a two-phase rewriting strategy that disguises harmful requests and then iteratively refines them. Extensive evaluation across commercial and open-source models (Llama-3.1-8B, GPT-4o mini, Qwen3-235B, Mistral-7B) reveals persistent vulnerabilities in current safety mechanisms, with multi-turn attacks consistently outperforming single-turn approaches. These findings demonstrate that alignment strategies optimized for single-turn interactions fail to maintain robustness across extended conversations, highlighting an urgent need for multi-turn-aware defenses.
format Preprint
id arxiv_https___arxiv_org_abs_2511_02376
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoAdv: Automated Adversarial Prompting for Multi-Turn Jailbreaking of Large Language Models
Reddy, Aashray
Zagula, Andrew
Saban, Nicholas
Computation and Language
Artificial Intelligence
Cryptography and Security
Machine Learning
Large Language Models (LLMs) remain vulnerable to jailbreaking attacks where adversarial prompts elicit harmful outputs. Yet most evaluations focus on single-turn interactions while real-world attacks unfold through adaptive multi-turn conversations. We present AutoAdv, a training-free framework for automated multi-turn jailbreaking that achieves an attack success rate of up to 95% on Llama-3.1-8B within six turns, a 24% improvement over single-turn baselines. AutoAdv uniquely combines three adaptive mechanisms: a pattern manager that learns from successful attacks to enhance future prompts, a temperature manager that dynamically adjusts sampling parameters based on failure modes, and a two-phase rewriting strategy that disguises harmful requests and then iteratively refines them. Extensive evaluation across commercial and open-source models (Llama-3.1-8B, GPT-4o mini, Qwen3-235B, Mistral-7B) reveals persistent vulnerabilities in current safety mechanisms, with multi-turn attacks consistently outperforming single-turn approaches. These findings demonstrate that alignment strategies optimized for single-turn interactions fail to maintain robustness across extended conversations, highlighting an urgent need for multi-turn-aware defenses.
title AutoAdv: Automated Adversarial Prompting for Multi-Turn Jailbreaking of Large Language Models
topic Computation and Language
Artificial Intelligence
Cryptography and Security
Machine Learning
url https://arxiv.org/abs/2511.02376