CoreML PoC Repository - Research Only

⚠️ Educational / Security Research Only
This repository demonstrates file-format vulnerabilities in Apple's coremltools pipeline (as of June 2026).

Files Included:

evil.safetensors → ACE via metadata deserialization
evil.gguf → Backdoor + output manipulation
evil.keras → Custom object RCE
evil.joblib → Joblib pickle RCE + credential stealer

Usage Warning:
Only load these in isolated VMs. Do not use in production or on real devices.

Research Context:
These PoCs were created to highlight gaps in model scanning and conversion safety in coremltools when handling .safetensors, .gguf, .keras, and .joblib formats.

Responsible Disclosure:
Reported to Apple Security. For research purposes only.

Model Card for Testing

Task: Malicious model demonstration
License: Research Only
Created for: coremltools vulnerability research

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support