
A deep dive into exchanging AWS cloud identities for internal X.509 certificates using double-claim corroboration
This article explains how Netflix securely bootstraps internal PKI identities for Apache Spark jobs running on managed AWS EMR. It details a robust workload attestation architecture that combines AWS STS pre-signed URLs with signed platform claims to issue short-lived certificates without relying on workload self-reporting.
Crucial reading for infrastructure, security, and platforms engineers looking to extend zero-trust internal service meshes to managed cloud or multi-tenant big data compute environments.
While Netflix relies on its private PKI system (Metatron) for service-to-service mTLS authentication, Apache Spark workloads on managed compute environments like Amazon EMR only launch with an AWS execution role, creating an identity gap with internal services.
The team established a 1:1 mapping between Data Project identities and dedicated AWS IAM roles. They bridged the identity systems by corroborating two independent claims: a control plane's signed workload metadata and a transferable proof of possession (a pre-signed AWS STS URL) generated by the Spark driver, which the Identity Service uses to issue Metatron X.509 certificates.
The solution safely provisions and continuously renews short-lived internal certificates for managed workloads, enabling secure resource access and consistent table-level audit/ACL enforcement without hitting AWS STS rate limits during massive Spark job fan-outs.
Trade-off
Distributing credentials from the driver to executors over secure Spark RPC avoids STS throttling but creates a second trust boundary within the application, making the driver a single point of credential distribution, while the corroboration logic introduces network round trips and strict cryptographic signing requirements.
The process of verifying a software workload's identity based on verifiable platform-specific attributes or proofs before granting access or issuing credentials.
A URL generated using a client's AWS credentials that grants temporary permission to perform a specific action, which can be shared and verified by third parties.
A security protocol in which both the client and the server authenticate each other using digital certificates, ensuring confidential and verified communications.




