大模型对齐与安全 - arXivDaily 专题

2305.06360 2024-10-28 cs.LG cs.AI 62%

Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomy

Thanveer Shaik, Xiaohui Tao, Haoran Xie, Lin Li, Xiaofeng Zhu, Qing Li

专题命中越狱攻击：trustworthy(abstract)；分类 cs.AI、cs.LG

Comments This work has been submitted to the IEEE for possible publication

2403.14774 2024-10-24 cs.CV cs.CL cs.CR cs.LG 62%

Few-Shot Adversarial Prompt Learning on Vision-Language Models

Yiwei Zhou, Xiaobo Xia, Zhiwei Lin, Bo Han, Tongliang Liu

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.LG

Comments NeurIPS 2024

2404.07242 2024-10-22 cs.CR cs.AI cs.CL 62%

Sandwich attack: Multi-language Mixture Adaptive Attack on LLMs

Bibek Upadhayay, Vahid Behzadan

专题命中越狱攻击：safety(abstract)；分类 cs.CL、cs.AI

2404.13968 2024-10-11 cs.CL cs.AI cs.CR 62%

Protecting Your LLMs with Information Bottleneck

Zichuan Liu, Zefan Wang, Linjie Xu, Jinyu Wang, Lei Song, Tianchun Wang, Chunlin Chen, Wei Cheng, Jiang Bian

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

Comments Accepted by Neural Information Processing Systems (NeurIPS 2024)

2405.08755 2024-10-10 cs.CR cs.AI cs.LG 62%

Distributed Threat Intelligence at the Edge Devices: A Large Language Model-Driven Approach

Syed Mhamudul Hasan, Alaa M. Alotaibi, Sajedul Talukder, Abdur R. Shahid

专题命中越狱攻击：trustworthy(abstract)；分类 cs.AI、cs.LG

Journal ref 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC)

2401.05949 2024-10-10 cs.CL cs.AI cs.CR 62%

Universal Vulnerabilities in Large Language Models: Backdoor Attacks for In-context Learning

Shuai Zhao, Meihuizi Jia, Luu Anh Tuan, Fengjun Pan, Jinming Wen

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.AI

2408.14866 2024-10-08 cs.CL cs.CR cs.LG 62%

Advancing Adversarial Suffix Transfer Learning on Aligned Large Language Models

Hongfu Liu, Yuxi Xie, Ye Wang, Michael Shieh

专题命中越狱攻击：safety(abstract)；分类 cs.CL、cs.LG

Comments Accepted to the EMNLP 2024

2409.19173 2024-10-01 cs.CL cs.AI 62%

HM3: Heterogeneous Multi-Class Model Merging

Stefan Hackmann

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2409.04822 2024-09-10 cs.CL cs.AI 62%

Exploring Straightforward Conversational Red-Teaming

George Kour, Naama Zwerdling, Marcel Zalmanovici, Ateret Anaby-Tavor, Ora Nova Fandina, Eitan Farchi

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.AI

2408.17235 2024-09-06 cs.CR cs.AI cs.LG 62%

AI-Driven Intrusion Detection Systems (IDS) on the ROAD Dataset: A Comparative Analysis for Automotive Controller Area Network (CAN)

Lorenzo Guerra, Linhan Xu, Paolo Bellavista, Thomas Chapuis, Guillaume Duc, Pavlo Mozharovskyi, Van-Tam Nguyen

专题命中越狱攻击：safety(abstract)；分类 cs.AI、cs.LG

2408.10107 2024-08-20 cs.LG cs.AI stat.ML 62%

Perturb-and-Compare Approach for Detecting Out-of-Distribution Samples in Constrained Access Environments

Heeyoung Lee, Hoyoon Byun, Changdae Oh, JinYeong Bak, Kyungwoo Song

专题命中越狱攻击：safety(abstract)；分类 cs.AI、cs.LG

Comments Accepted to European Conference on Artificial Intelligence (ECAI) 2024

2407.15399 2024-07-23 cs.CL cs.AI cs.CR 62%

Imposter.AI: Adversarial Attacks with Hidden Intentions towards Aligned Large Language Models

Xiao Liu, Liangzhi Li, Tong Xiang, Fuying Ye, Lu Wei, Wangyue Li, Noa Garcia

专题命中越狱攻击：safety(abstract)；分类 cs.CL、cs.AI

2407.13796 2024-07-22 cs.CR cs.AI cs.CL 62%

Continuous Embedding Attacks via Clipped Inputs in Jailbreaking Large Language Models

Zihao Xu, Yi Liu, Gelei Deng, Kailong Wang, Yuekang Li, Ling Shi, Stjepan Picek

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2407.12822 2024-07-19 cs.CL cs.AI 62%

Lightweight Large Language Model for Medication Enquiry: Med-Pal

Kabilan Elangovan, Jasmine Chiat Ling Ong, Liyuan Jin, Benjamin Jun Jie Seng, Yu Heng Kwan, Lit Soo Tan, Ryan Jian Zhong, Justina Koi Li Ma, YuHe Ke, Nan Liu, Kathleen M Giacomini, Daniel Shu Wei Ting

专题命中越狱攻击：safety(abstract)；分类 cs.CL、cs.AI

2407.03160 2024-07-04 cs.CR cs.CL cs.LG 62%

SOS! Soft Prompt Attack Against Open-Source Large Language Models

Ziqing Yang, Michael Backes, Yang Zhang, Ahmed Salem

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.LG

2406.15518 2024-06-25 cs.CL cs.LG 62%

Steering Without Side Effects: Improving Post-Deployment Control of Language Models

Asa Cooper Stickland, Alexander Lyzhov, Jacob Pfau, Salsabila Mahdi, Samuel R. Bowman

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.LG

2402.14857 2024-06-21 cs.CL cs.AI cs.CR 62%

Is the System Message Really Important to Jailbreaks in Large Language Models?

Xiaotian Zou, Yongkang Chen, Ke Li

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

Comments 13 pages,3 figures

2406.00240 2024-06-04 cs.LG cs.CL cs.CR 62%

Exploring Vulnerabilities and Protections in Large Language Models: A Survey

Frank Weizhen Liu, Chenhui Hu

专题命中越狱攻击：prompt injection(abstract)；分类 cs.CL、cs.LG

2405.15245 2024-05-27 cs.LG cs.AI 62%

Cooperative Backdoor Attack in Decentralized Reinforcement Learning with Theoretical Guarantee

Mengtong Gao, Yifei Zou, Zuyuan Zhang, Xiuzhen Cheng, Dongxiao Yu

专题命中越狱攻击：safety(abstract)；分类 cs.AI、cs.LG

2308.07308 2024-05-03 cs.CL cs.AI 62%

LLM Self Defense: By Self Examination, LLMs Know They Are Being Tricked

Mansi Phute, Alec Helbling, Matthew Hull, ShengYun Peng, Sebastian Szyller, Cory Cornelius, Duen Horng Chau

专题命中越狱攻击：safety(abstract)；分类 cs.CL、cs.AI

2404.02406 2024-04-04 cs.CR cs.AI cs.CL 62%

Exploring Backdoor Vulnerabilities of Chat Models

Yunzhuo Hao, Wenkai Yang, Yankai Lin

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.AI

Comments Code and data are available at https://github.com/hychaochao/Chat-Models-Backdoor-Attacking

2404.02067 2024-04-03 cs.CV cs.AI cs.LG 62%

Red-Teaming Segment Anything Model

Krzysztof Jankowski, Bartlomiej Sobieski, Mateusz Kwiatkowski, Jakub Szulc, Michal Janik, Hubert Baniecki, Przemyslaw Biecek

专题命中越狱攻击：safety(abstract)；分类 cs.AI、cs.LG

Comments CVPR 2024 - The 4th Workshop of Adversarial Machine Learning on Computer Vision: Robustness of Foundation Models

2403.12171 2024-03-20 cs.CL cs.AI 62%

EasyJailbreak: A Unified Framework for Jailbreaking Large Language Models

Weikang Zhou, Xiao Wang, Limao Xiong, Han Xia, Yingshuang Gu, Mingxu Chai, Fukang Zhu, Caishuang Huang, Shihan Dou, Zhiheng Xi, Rui Zheng, Songyang Gao, Yicheng Zou, Hang Yan, Yifan Le, Ruohui Wang, Lijun Li, Jing Shao, Tao Gui, Qi Zhang, Xuanjing Huang

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2211.14769 2024-03-19 cs.AI cs.CL cs.CR cs.CV 62%

Navigation as Attackers Wish? Towards Building Robust Embodied Agents under Federated Learning

Yunchao Zhang, Zonglin Di, Kaiwen Zhou, Cihang Xie, Xin Eric Wang

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.AI

2305.13860 2024-03-12 cs.SE cs.AI cs.CL 62%

Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study

Yi Liu, Gelei Deng, Zhengzi Xu, Yuekang Li, Yaowen Zheng, Ying Zhang, Lida Zhao, Tianwei Zhang, Kailong Wang, Yang Liu

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2311.16119 2024-03-05 cs.CR cs.AI cs.CL 62%

Ignore This Title and HackAPrompt: Exposing Systemic Vulnerabilities of LLMs through a Global Scale Prompt Hacking Competition

Sander Schulhoff, Jeremy Pinto, Anaum Khan, Louis-François Bouchard, Chenglei Si, Svetlina Anati, Valen Tagliabue, Anson Liu Kost, Christopher Carnahan, Jordan Boyd-Graber

专题命中越狱攻击：prompt injection(abstract)；分类 cs.CL、cs.AI

Comments 34 pages, 8 figures Codebase: https://github.com/PromptLabs/hackaprompt Dataset: https://huggingface.co/datasets/hackaprompt/hackaprompt-dataset/blob/main/README.md Playground: https://huggingface.co/spaces/hackaprompt/playground

2402.15690 2024-02-27 cs.CL cs.AI 62%

Foot In The Door: Understanding Large Language Model Jailbreaking via Cognitive Psychology

Zhenhua Wang, Wei Xie, Baosheng Wang, Enze Wang, Zhiwen Gui, Shuoyoucheng Ma, Kai Chen

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2402.15570 2024-02-27 cs.CR cs.AI cs.CL 62%

Fast Adversarial Attacks on Language Models In One GPU Minute

Vinu Sankar Sadasivan, Shoumik Saha, Gaurang Sriramanan, Priyatham Kattakinda, Atoosa Chegini, Soheil Feizi

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.AI

2402.14020 2024-02-22 cs.LG cs.CL cs.CR 62%

Coercing LLMs to do and reveal (almost) anything

Jonas Geiping, Alex Stein, Manli Shu, Khalid Saifullah, Yuxin Wen, Tom Goldstein

专题命中越狱攻击：jailbreak(abstract)；分类 cs.CL、cs.LG

Comments 32 pages. Implementation available at https://github.com/JonasGeiping/carving

2311.11509 2024-02-20 cs.CL cs.LG 62%

Token-Level Adversarial Prompt Detection Based on Perplexity Measures and Contextual Information

Zhengmian Hu, Gang Wu, Saayan Mitra, Ruiyi Zhang, Tong Sun, Heng Huang, Viswanathan Swaminathan

专题命中越狱攻击：alignment(abstract)；分类 cs.CL、cs.LG

1. 越狱攻击 1717 篇

Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomy

Few-Shot Adversarial Prompt Learning on Vision-Language Models

Sandwich attack: Multi-language Mixture Adaptive Attack on LLMs

Protecting Your LLMs with Information Bottleneck

Distributed Threat Intelligence at the Edge Devices: A Large Language Model-Driven Approach

Universal Vulnerabilities in Large Language Models: Backdoor Attacks for In-context Learning

Advancing Adversarial Suffix Transfer Learning on Aligned Large Language Models

HM3: Heterogeneous Multi-Class Model Merging

Exploring Straightforward Conversational Red-Teaming

AI-Driven Intrusion Detection Systems (IDS) on the ROAD Dataset: A Comparative Analysis for Automotive Controller Area Network (CAN)

Perturb-and-Compare Approach for Detecting Out-of-Distribution Samples in Constrained Access Environments

Imposter.AI: Adversarial Attacks with Hidden Intentions towards Aligned Large Language Models

Continuous Embedding Attacks via Clipped Inputs in Jailbreaking Large Language Models

Lightweight Large Language Model for Medication Enquiry: Med-Pal

SOS! Soft Prompt Attack Against Open-Source Large Language Models

Steering Without Side Effects: Improving Post-Deployment Control of Language Models

Is the System Message Really Important to Jailbreaks in Large Language Models?

Exploring Vulnerabilities and Protections in Large Language Models: A Survey

Cooperative Backdoor Attack in Decentralized Reinforcement Learning with Theoretical Guarantee

LLM Self Defense: By Self Examination, LLMs Know They Are Being Tricked

Exploring Backdoor Vulnerabilities of Chat Models

Red-Teaming Segment Anything Model

EasyJailbreak: A Unified Framework for Jailbreaking Large Language Models

Navigation as Attackers Wish? Towards Building Robust Embodied Agents under Federated Learning

Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study

Ignore This Title and HackAPrompt: Exposing Systemic Vulnerabilities of LLMs through a Global Scale Prompt Hacking Competition

Foot In The Door: Understanding Large Language Model Jailbreaking via Cognitive Psychology

Fast Adversarial Attacks on Language Models In One GPU Minute

Coercing LLMs to do and reveal (almost) anything

Token-Level Adversarial Prompt Detection Based on Perplexity Measures and Contextual Information