Adversarial Search Engine Optimization for Large Language Models

arXiv:2406.18382v1 Announce Type: new
Abstract: Large Language Models (LLMs) are increasingly used in applications where the model selects from competing third-party content, such as in LLM-powered search engines or chatbot plugins. In this paper, we introduce Preference Manipulation Attacks, a new class of attacks that manipulate an LLM's selections to favor the attacker. We demonstrate that carefully crafted website content or plugin documentations can trick an LLM to promote the attacker products and discredit competitors, thereby increasing user traffic and …

adversarial applications arxiv attacks chatbot class cs.cr cs.lg engine language language models large llm llms manipulation optimization party plugins search search engine search engines third third-party