Securely Executing Python Machine Learning Models with Distroless Images at ING

Explore how to securely execute Python machine learning models in a production environment using distroless images at a major bank. This EuroPython 2019 talk delves into potential security risks of running Docker containers in production, methods for enhancing container security through minimal Docker images for Python, and practical applications of these images for serving machine learning models at ING. Gain insights into conducting security analyses, implementing distroless images, and utilizing PI installers to create secure model-serving Docker images. Suitable for those with basic knowledge of Docker and security concepts, this presentation aims to equip you with strategies for mitigating risks and optimizing the deployment of Python models in high-compliance environments.

Introduction
Machine learning in production
Random forest classifier
Security analysis
Distroless images
PI installer
Summary
Questions