paper-with-me

홈 › Papers

BadGPT: Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT

2023-02-21 · Jiawen Shi, Yixin Liu, Pan Zhou, Lichao Sun

Recently, ChatGPT has gained significant attention in research due to its ability to interact with humans effectively. The core idea behind this model is reinforcement learning (RL) fine-tuning, a new paradigm that allows language models to align with human preferences, i.e., InstructGPT. In this study, we propose BadGPT, the first backdoor attack against RL fine-tuning in language models. By injecting a backdoor into the reward model, the language model can be compromised during the fine-tuning stage. Our initial experiments on movie reviews, i.e., IMDB, demonstrate that an attacker can manipulate the generated text through BadGPT.

📄 PDF Abstract BibTeX arXiv:2304.12298

Code (0)

등록된 구현이 없습니다.

Tasks

Backdoor AttackLanguage ModelingLanguage Modellingreinforcement-learningReinforcement Learning (RL)

Methods 이 논문이 사용한 방법론

ALIGN In the ALIGN method, visual and language representations are jointly trained from noisy image alt-text data. The image and text encoders are learned via contrastive loss…

Similar Papers 제목 키워드 기반

An Investigation into Misuse of Java Security APIs by Large Language Models

2024-04-04 · Zahra Mousavi, Chadni Islam, Kristen Moore, Alsharif Abuadbba 외

The increasing trend of using Large Language Models (LLMs) for code generation raises the question of their capability to generate trustworthy code. While many researchers are exploring the utility of code generation for…

Code Generation

Backdooring Convolutional Neural Networks via Targeted Weight Perturbations

2018-12-07 · Jacob Dumford, Walter Scheirer

We present a new type of backdoor attack that exploits a vulnerability of convolutional neural networks (CNNs) that has been previously unstudied. In particular, we examine the application of facial recognition. Deep lea…

Backdoor Attack

BadSAM: Exploring Security Vulnerabilities of SAM via Backdoor Attacks

2023-05-05 · Zihan Guan, Mengxuan Hu, Zhongliang Zhou, Jielu Zhang 외

Recently, the Segment Anything Model (SAM) has gained significant attention as an image segmentation foundation model due to its strong performance on various downstream tasks. However, it has been found that SAM does no…

Backdoor AttackImage SegmentationSegmentationSemantic Segmentation

Exploring Vulnerabilities and Protections in Large Language Models: A Survey

2024-06-01 · Frank Weizhen Liu, Chenhui Hu

As Large Language Models (LLMs) increasingly become key components in various AI applications, understanding their security vulnerabilities and the effectiveness of defense mechanisms is crucial. This survey examines the…

Data PoisoningSurvey

Exploring Backdoor Vulnerabilities of Chat Models

2024-04-03 · Yunzhuo Hao, Wenkai Yang, Yankai Lin

Recent researches have shown that Large Language Models (LLMs) are susceptible to a security threat known as Backdoor Attack. The backdoored model will behave well in normal cases but exhibit malicious behaviours on inpu…

Backdoor Attack