# Protectai

# Online Protect AI Scanner

Search for the name of the model before downloading
- https://protectai.com/insights

# Protect AI Scanner CLI

https://protectai.com/

https://github.com/protectai/modelscan

https://github.com/protectai/modelscan/blob/main/docs/attack-1.png

What Models and Frameworks Are Supported?

https://github.com/protectai/modelscan?tab=readme-ov-file#what-models-and-frameworks-are-supported

This will be expanding continually, so look out for changes in our release notes.

At present, ModelScan supports any Pickle derived format and many others:

ML Library	API	Serialization Format	modelscan support
Pytorch	torch.save() and torch.load()	Pickle	Yes
Tensorflow	tf.saved_model.save()	Protocol Buffer	Yes
Keras	keras.models.save(save_format= 'h5')	HD5 (Hierarchical Data Format)	Yes
	keras.models.save(save_format= 'keras')	Keras V3 (Hierarchical Data Format)	Yes
Classic ML Libraries (Sklearn, XGBoost etc.)	pickle.dump(), dill.dump(), joblib.dump(), cloudpickle.dump()	Pickle, Cloudpickle, Dill, Joblib	Yes

# Command line

First download the file to scan locally, you cant scan it on online repo
- Harmless model
  - https://huggingface.co/ykilcher/totally-harmless-model/blob/main/pytorch_model.bin
Commands

pip install modelscan

modelscan -p /path/to/model_file.pkl
Expected Result

# For later

https://github.com/protectai#protect-ai-oss

Project	Description
ModelScan	🔍 ML Model Security Scanner
AI Exploits	🗡️ Collection of AI/ML Exploits
LLM Guard	🛡️ Security Toolkit for LLM Interactions
rebuff	💉 LLM Prompt Injection Attack Detection
NB Defense	📓 Jupyter Notebooks Security
Vulnhuntr	🏹 Autonomous AI-Discovered 0Day Tool