Seldon Deployment Token
Service Name: Seldon Core
Service Description: Seldon Core is an open-source platform for deploying machine learning models on Kubernetes. It provides tools for packaging, deploying, monitoring, and managing machine learning models at scale.
Service Address: https://www.seldon.io/
Validation Type: -
IP Allow list: IP restrictions can be configured at the Kubernetes cluster level using network policies.
Secret Access Scope: Grants access to deploy, manage, and interact with machine learning models deployed on Seldon Core within a Kubernetes cluster.
Secret Revokement URL: Does not exist (managed through Kubernetes secrets management)
Secret Example: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzdWIiOiJzZWxkb24tZGVwbG95bWVudCIsI m5hbWUiOiJTZWxkb24gRGVwbG95bWVudCBUb2tlbiIsImlhdCI6MTUxNjIzOTAyMn0.EXAMPLE_ TOKEN_SIGNATURE
Suspicious Activity Investigation Instructions:
- Review Kubernetes logs for unusual deployment or model serving activities
- Check for unexpected model deployments or modifications to existing deployments
- Monitor API request patterns to deployed models for anomalies
- Examine Kubernetes audit logs for unauthorized access attempts
- Verify resource utilization patterns for signs of abuse or misuse
Mitigation Instructions:
- Rotate the compromised deployment token by creating a new Kubernetes
secret
- Delete the compromised token from the Kubernetes cluster
- Update any applications or services using the token with the new credentials
- Review and tighten RBAC permissions for the service account associated with
the token
- Implement network policies to restrict access to the Seldon deployment
endpoints
- Enable audit logging for the Kubernetes cluster to monitor future access
- Consider implementing additional authentication mechanisms such as mTLS