Instructions to use hacnho/keras-melspectrogram-maxfreq-trigger-poc with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Keras
How to use hacnho/keras-melspectrogram-maxfreq-trigger-poc with Keras:
# Available backend options are: "jax", "torch", "tensorflow". import os os.environ["KERAS_BACKEND"] = "jax" import keras model = keras.saving.load_model("hf://hacnho/keras-melspectrogram-maxfreq-trigger-poc") - Notebooks
- Google Colab
- Kaggle
Keras MelSpectrogram max_freq trigger PoC
This is a benign security research PoC for the Huntr Model File Vulnerability program.
Files:
control_melspectrogram_maxfreq_8000.kerasmalicious_melspectrogram_maxfreq_1000.kerasreproduce.pyrequirements.txt
The malicious .keras file differs from the control by one serialized config field:
config.json:MelSpectrogram(name=mel_gate).max_freq
control: 8000.0
malicious: 1000.0
Reproduction:
python -m pip install -r requirements.txt
python reproduce.py control_melspectrogram_maxfreq_8000.keras malicious_melspectrogram_maxfreq_1000.keras
Expected result:
benign preds: control [1, 1], malicious [1, 1]
trigger preds: control [0, 1], malicious [1, 1]
target bin: 76.72 -> near 0
modelscan: No issues found
Tested with keras==3.15.0, tensorflow==2.19.0, and modelscan==0.8.8.
Public files:
- Control: https://huggingface.co/hacnho/keras-melspectrogram-maxfreq-trigger-poc/resolve/main/control_melspectrogram_maxfreq_8000.keras
- Malicious: https://huggingface.co/hacnho/keras-melspectrogram-maxfreq-trigger-poc/resolve/main/malicious_melspectrogram_maxfreq_1000.keras
- Reproducer: https://huggingface.co/hacnho/keras-melspectrogram-maxfreq-trigger-poc/resolve/main/reproduce.py
- Downloads last month
- 101
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support