Article
2 min read
Andy Rossiter
  • Global SVP of Google Cloud at Endava

We've established the security paradox, defined the challenges in securing data and models, and introduced the two key solutions: Wiz (for proactive risk reduction) and Google SecOps (for real-time threat detection). 

 

Now, let's look at the powerful synergy of these tools and how they combine to secure the entire MLOps lifecycle on Google Vertex AI. 

 

Securing the AI lifecycle: a joint  defence

 

The secure adoption of Vertex AI is achieved by segmenting the security needs into two distinct phases, each handled by the optimal tool:  

 

  1. 1. Pre-deployment: establishing the secure baseline (Wiz's domain)

  2.  

This phase focuses on ensuring the foundation is secure before any model is trained or deployed.  

 

  • Data security: Wiz automatically scans GCS buckets or BigQuery tables used for training data. It flags misconfigurations like excessive permissions or public access, ensuring sensitive data is not over-exposed to Vertex AI service accounts or users.  
  • Infrastructure hygiene: Wiz identifies and prioritizes vulnerabilities (CVEs) in the container images, GKE clusters or Compute instances that form the backbone of your Vertex AI Workbench and custom training jobs.  
  • Secure by design: By integrating Wiz scans into the CI/CD pipeline, security issues are flagged and fixed instantly, accelerating MLOps without compromising security integrity.  

 

  1. 2. Post-deployment: continuous monitoring and response (Google SecOps' domain) 

  2.  

This phase is about constant vigilance and rapid response to active threats on running AI workloads.  

 

  • Threat detection for running models: Chronicle Security Operations ingests high-fidelity logs from the Vertex AI API and underlying Google Cloud infrastructure. It can detect anomalous activity that indicates a compromise, such as:  
      • An AI service account suddenly accessing new or unrelated data sources.  
      • Suspicious API calls to a deployed Vertex AI Endpoint attempting to test for adversarial weaknesses.  
      • Large volumes of data being moved out of a training data bucket (potential exfiltration).  
  • Informed response: If a threat is detected, Chronicle centralizes the data for rapid investigation. Mandiant intelligence informs the security team whether the activity matches known attack campaigns, allowing for a precise and surgical response.  

 

The ultimate business outcome: confidence and speed

 

By working together, Wiz and Google SecOps solve the security paradox and deliver tangible business benefits:  
 

  1. 1. Accelerated time-to-value: Security moves from being a bottleneck to a feature. Developers and data scientists can use Vertex AI more freely and frequently, knowing their environments are continuously pre-vetted and monitored.  
  2.  
  1. 2. Continuous compliance: With full, continuous visibility into the security posture of all AI components and auditable logs of all activity, you can easily prove compliance with industry regulations and internal governance policies.  
  2.  
  1. 3. Future-proof protection: Your organization gains a unified defence against both known vulnerabilities in the cloud stack and unknown adversarial threats targeting the AI layer. 
     

The secure path to AI innovation is no longer a trade-off. By establishing a unified security architecture powered by Wiz and Google SecOps on Vertex AI, your organization can build, deploy and innovate with confidence.  

 

Ready to accelerate your secure AI journey?  

 

Contact our team for a personalized assessment of your current security posture and a demo of the Wiz and Google SecOps integration on your Google Cloud environment.