AI Security with MLSecOps

In the previous chapter, we discussed our secure-by-design AI methodology that can help us identify threats relevant to our solution and apply appropriate mitigations and controls. We highlighted the need for MLSecOps to be a fundamental enabler of AI security. In this chapter, we will discuss in more detail, with practical demonstrations, why MLSecOps are so essential and provide a practical exploration of the concepts.

We will cover the following topics:

• How modern AI trends are accelerating the imperative of MLSecOps
• MLSecOps workflows to evolve from DevSecOps and MLOps to MLSecOps
• Building a simple MLSecOPs platform with Jenkins and MLFlow
• MLSecOps in action with Jenkins, MLFlow, and our sample solution
• Integrating MLSecOps with notebook-based interactive workflows
• The MLSecOps aspects of LLMs and the role of LLMOps
• Advanced use of MLSecOps with ML SBOMs

Let’s start with a refresher on the critical concepts...