Maliciously crafted fonts can hide adversarial prompts from human eyes while LLMs read and obey them, enabling content manipulation and data exfiltration via MCP tools.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Invisible Prompts, Visible Threats: Malicious Font Injection in External Resources for Large Language Models
Maliciously crafted fonts can hide adversarial prompts from human eyes while LLMs read and obey them, enabling content manipulation and data exfiltration via MCP tools.