Which statement about Cilium's identity-based security model is correct?

Technical explanation

Cilium derives a workload’s security identity from its security-relevant labels. Network policies then refer to workload characteristics such as application, role, environment, namespace, or service account rather than depending exclusively on transient pod IP addresses. The identity is associated with traffic in the Cilium datapath and validated when policy is enforced. This makes B the accurate description.

The identity is not limited to a single pod. Endpoints that have the same set of identity-relevant labels can share the same numeric security identity, including endpoints located on different cluster nodes. This reduces policy-map growth and allows policy to scale with logical application groups rather than with the number of pod addresses. Namespace information is normally among the labels used to derive identity, but that does not make an identity inherently “tied to a single namespace” as option A states.

Options C and D invert Cilium’s design. IP addresses remain necessary for packet delivery, but they are not the primary security identifier for Cilium-managed workloads. Pods can be recreated and assigned new addresses while retaining the same relevant labels and therefore the same security intent. Decoupling identity from addressing is precisely what improves scalability and operational stability.

Official references

Cilium Terminology and Identity ; Introduction to Cilium and Hubble .

Study Guide topic: Architecture.