Prompt Overflow: What the Guardrail Inspects Is Not What the Model Infers

Guardrail models a.k.a. safety checkers are widely deployed to screen user inputs before they reach large language models LLMs, serving as a primary defense against prompt injection attacks. Due to strict context constraints, these models handle overlength prompts through truncation or...

CVE-2023-21260

In notification access permission dialog box, malicious application can embedded a very long service label that overflow the original user prompt and possibly contains mis-leading information to be appeared as a system message for user confirmation...