Token Level Attack

Token-level attacks aim to manipulate large language models (LLMs) by subtly altering individual words or sub-word units within prompts, causing the model to generate undesirable or harmful outputs. Current research focuses on developing increasingly sophisticated attack methods, employing techniques like Markov decision processes and gradient-based optimization to find effective adversarial perturbations, often within a black-box setting. These attacks highlight vulnerabilities in LLMs and drive the development of more robust models and defenses, impacting the safety and reliability of AI systems across various applications. The effectiveness of these attacks varies across different LLMs, suggesting ongoing improvements in model robustness are needed.

Papers

October 9, 2024

Break the Visual Perception: Adversarial Attacks Targeting Encoded Visual Tokens of Large Vision-Language Models
Yubo Wang, Chaohu Liu, Yanqiu Qu, Haoyu Cao, Deqiang Jiang, Linli Xu
Adversarial Attack Large Vision Language Model Visual Perception Adversarial Image Visual Token BREAK for Make Token Level Attack

August 11, 2024

Kov: Transferable and Naturalistic Black-Box LLM Attacks using Markov Decision Processes and Tree Search
Robert J. Moss
Adversarial Attack Markov Decision Process Tree Search Black Box LLM Word Level Attack LLM Attack Token Level Attack

August 7, 2024

EnJa: Ensemble Jailbreak on Large Language Models
Jiahao Zhang, Zilong Wang, Ruofan Wang, Xingjun Ma, Yu-Gang Jiang
Jailbreak Attack Token Level Attack

May 15, 2024

Efficient LLM Jailbreak via Adaptive Dense-to-sparse Constrained Optimization
Kai Hu, Weichen Yu, Tianjun Yao, Xiang Li, Wenhe Liu, Lijun Yu, Yining Li, Kai Chen, Zhiqiang Shen, Matt Fredrikson
Large Language Model Jailbreak Attack Token Level Sparsity Constraint Token Level Attack

May 7, 2024

Revisiting Character-level Adversarial Attacks for Language Models
Elias Abad Rocamora, Yongtao Wu, Fanghui Liu, Grigorios G. Chrysos, Volkan Cevher
Adversarial Attack Adversarial Example Token Level Attack

February 15, 2023

COVID-VTS: Fact Extraction and Verification on Short Video Platforms
Fuxiao Liu, Yaser Yacoob, Abhinav Shrivastava
Covid 19 Verification Task Multimodal Information Short Video Fact Extraction Token Level Attack

May 23, 2022

Learning to Ignore Adversarial Attacks
Yiming Zhang, Yangqiaoyu Zhou, Samuel Carton, Chenhao Tan
LeArning Abstract Adversarial Attack Adversarial Example Natural Language Processing Model Adversarial Input Token Level Attack