attack arXiv Aug 25, 2025 · Aug 2025
Kyohei Shiomi, Zhuotao Lian, Toru Nakanishi et al. · Hiroshima University
Attacks LLM-powered game NPCs via prompt injection to extract developer-embedded secrets from system prompts
Prompt Injection Sensitive Information Disclosure nlp
Large Language Models (LLMs) are increasingly used to generate dynamic dialogue for game NPCs. However, their integration raises new security concerns. In this study, we examine whether adversarial prompt injection can cause LLM-based NPCs to reveal hidden background secrets that are meant to remain undisclosed.
llm Hiroshima University
ML Security Papers