MENU
Your organization is building a real-time recommendation engine using ML models that process live user activity data stored in BigQuery and Cloud Storage. Each new model developed is saved to Artifact Registry. This new system deploys models to Google Kubernetes Engine and uses Pub/Sub for message queues. Recent industry news has been reporting attacks exploiting ML model supply chains. You need to enhance the security in this serverless architecture, specifically against risks to the development and deployment pipeline. What should you do?
To enhance the security of your machine learning (ML) model supply chain within a serverless architecture, it's crucial to implement measures that protect both the development and deployment pipelines.
Option A: While limiting external dependencies and rotating encryption keys are good security practices, they do not directly address the risks associated with the ML model supply chain.
Option B: Implementing container image vulnerability scanning during development and pre-deployment helps identify and mitigate known vulnerabilities in your container images. Enforcing Binary Authorization ensures that only trusted and verified images are deployed in your environment. This combination directly strengthens the security of the ML model supply chain by validating the integrity of container images before deployment.
Option C: Sanitizing training data and applying role-based access controls are important security practices but do not specifically safeguard the deployment pipeline against compromised container images.
Option D: While strict firewall rules and intrusion detection systems enhance network security, they do not specifically address vulnerabilities within the container images or the deployment process.
Therefore, Option B is the most effective approach, as it directly addresses the security of the development and deployment pipeline by ensuring that only vetted and secure container images are used in your environment.