Tersorflow Remote Code Execution with Malicious Model

Repository for the scripts presented in the blogpost: https://mastersplinter.work/research/tensorflow-rce/

Files

• exploit.py: used to create a simple malicious model with a reverse shell
• inject.py: injects the malicious layer in a legitimate model
• model.py: simulate the usage of load_mode on a malicious model
• detector.py: detects a malicious Lambda in a .h5 model

Disclaimer

The contents of this repository are exclusively for research and entertainment purposes.