
Summary
This rule detects the first-time use of commonly abused base and debugging container images in Kubernetes pod creation requests. It monitors Kubernetes audit logs for allowed pod creation events where the pod spec's container images match patterns for busybox, alpine, ubuntu/debian, netshoot, network-multitool, as well as container runtime/build images like dind, podman, buildkit, buildah, and kaniko. By surfacing new image usage per source, user agent, and user within a 5-day lookback window, it helps identify attacker tradecraft such as short-lived or interactive containers used for reconnaissance, payload staging, or execution, and potential attempts to bypass image registry controls or run nested containers on the host. The rule is categorized as a New Terms detection and highlights first occurrences (per source, user agent, and user) of these images, using fields like source.ip, user_agent.original, and user.name within history_window_start now-5d. It aligns with MITRE ATT&CK techniques related to deploying and configuring containers (T1609 Container Administration Command, T1610 Deploy Container) and on-host image building (T1612 Build Image on Host), under Execution and Defense Evasion tactics. The detection targets Kubernetes audit logs (kubernetes.audit_logs), specifically the ResponseComplete stage for create operations on pods, where the authorization decision is allow and the image field matches the specified image patterns.
Categories
- Kubernetes
- Containers
- Cloud
Data Sources
- Container
- Image
ATT&CK Techniques
- T1609
- T1610
- T1612
Created: 2026-05-04