Machine Learning Model Threat Categories

TensorFlow SavedModel with malicious operators

Zip-based model attempting ZipSlip path traversal

TensorFlow SavedModel with Lambda Layers containing malicious code

PyTorch model with embedded malicious code

Model does not support code execution on load but contains suspicious code strings

PyTorch model using getattr maliciously

Pickle-based model with embedded malicious code

GGUF model attempting template injection for arbitrary code execution

ONNX model with architectural backdoor causing unexpected outputs

Keras model with Lambda Layers containing malicious code

Keras model with Custom Layers calling malicious functions

Pickle-based model using getattr maliciously

TensorFlow H5 model with Lambda Layers containing malicious code