user-model communication pipline的安全风险分类,重点关注prompt-based attack on LLM。根据目标和攻击类型分类,并使用具体示例对每种类型的影响进行展示。

Introduction

Untitled

the user, the model, and a third party

Taxonomy

Attack Target

CIA Triad

Shield 提出使用integrity, source identification, attack detectability, and utility preservation几种属性进行防御的方法Shield。

User

Untitled