SPML: A DSL for Defending Language Models Against Prompt Attacks

arXiv:2402.11755v1 Announce Type: cross
Abstract: Large language models (LLMs) have profoundly transformed natural language applications, with a growing reliance on instruction-based definitions for designing chatbots. However, post-deployment the chatbot definitions are fixed and are vulnerable to attacks by malicious users, emphasizing the need to prevent unethical applications and financial losses. Existing studies explore user prompts' impact on LLM-based chatbots, yet practical methods to contain attacks on application-specific chatbots remain unexplored. This paper presents System Prompt Meta Language (SPML), a domain-specific …

arxiv attacks cs.cl cs.cr cs.lg cs.pl defending dsl language language models prompt