Affected Systems
Google Cloud Vertex AI SDK for Python. Specific affected versions not disclosed. Impacts organizations using the SDK to upload and deploy machine learning models to Google Cloud's serving infrastructure.
Exploitation Status
No active exploitation observed in the wild. Vulnerability discovered and reported by Palo Alto Networks Unit 42 through Google's bug bounty program. Proof-of-concept technique ("Pickle in the Middle") has been documented by researchers.
Business Impact
Attackers without project access can intercept and manipulate ML model uploads to inject malicious code that executes within Google's serving infrastructure. This could lead to data exfiltration, model poisoning, unauthorized access to cloud resources, or supply chain compromise. Organizations using Vertex AI SDK for model deployment face risk of code execution in production ML pipelines. No CVE assigned yet; severity rated high by reporter.
Urgency
🟠Within 24 hours
Recommended Actions
- Update Google Cloud Vertex AI SDK for Python to the latest patched version immediately via pip (pip install --upgrade google-cloud-aiplatform)
- Review recent model upload activity in Vertex AI logs for anomalous behavior or unexpected model modifications
- Implement network-level controls to restrict model upload endpoints to trusted IP ranges and enforce mutual TLS authentication
- Audit IAM permissions for Vertex AI resources to ensure least-privilege access and enable Cloud Audit Logs for model deployment events
- Consider implementing model integrity verification (cryptographic signatures) before deployment until patch is confirmed applied