Hi

I just posted a blog + Notebook covering two vulnerabilities in ML pipelines, how to exploit them, and how to defend using -among other strategies,- secure serialization with the ONNX format.

I hope this is useful for someone in this community

Thanks!