file-protection
Installation
SKILL.md
LLM File Protection
Security Requirement
Large Language Models are composed of files (weights, configuration, tokenizer, etc.) loaded by the inference engine. These files MUST be protected against unauthorized access and modification.
Risks of Unprotected Model Files
| Risk | Impact |
|---|---|
| Unauthorized access | Model theft, intellectual property loss |
| Unauthorized modification | Backdoored or poisoned model served to users |
| Tampering | Modified weights producing biased, harmful, or incorrect outputs |
| Data extraction | Extracting sensitive training data embedded in model weights |