BaseIncomplete
CWE-1427Improper Neutralization of Input Used for LLM Prompting
Category: other
Description
The product uses externally-provided data to build prompts provided to
large language models (LLMs), but the way these prompts are constructed
causes the LLM to fail to distinguish between user-supplied inputs and
developer provided system directives.
Common consequences· 4
- Confidentiality / Integrity / Availability — Execute Unauthorized Code or Commands, Varies by Context
- Confidentiality — Read Application Data
- Integrity — Modify Application Data, Execute Unauthorized Code or Commands
- Access Control — Read Application Data, Modify Application Data, Gain Privileges or Assume Identity
Potential mitigations· 5
- [Architecture and Design]
- [Implementation]
- [Architecture and Design]
- [Implementation]
- [Installation, Operation]
References
Related by meaning· 6
Nearest entities by semantic similarity across the cs-graph corpus.