CVE-2026-27173
JWT Token Exposure in Kubernetes Executors via Read-Only Pod Access
Publication date: 2026-05-19
Last updated on: 2026-05-19
Assigner: Apache Software Foundation
Description
Description
CVSS Scores
EPSS Scores
| Probability: | |
| Percentile: |
Meta Information
Affected Vendors & Products
| Vendor | Product | Version / Range |
|---|---|---|
| apache | airflow | 10.17.0 |
| apache | apache-airflow-providers-cncf-kubernetes | to 10.17.0 (exc) |
Helpful Resources
Exploitability
| CWE ID | Description |
|---|---|
| CWE-538 | The product places sensitive information into files or directories that are accessible to actors who are allowed to have access to the files, but not to the sensitive information. |
Attack-Flow Graph
AI Powered Q&A
Can you explain this vulnerability to me?
This vulnerability involves the exposure of JWT tokens used by workers in Kubernetes Executors within Apache Airflow. Users who have only read-only access to Kubernetes Pods can access these tokens. Because of this exposure, such users could perform actions that are normally restricted to running tasks via the Task SDK. This potentially allows them to modify the state of the Airflow database related to tasks.
How can this vulnerability impact me? :
If you have users with read-only access to Kubernetes Pods, this vulnerability could allow them to escalate their privileges by using exposed JWT tokens. They could perform actions that should only be available to running tasks, such as modifying the state of the Airflow database for tasks. This could lead to unauthorized changes in task execution or data integrity issues within your Airflow environment.
What immediate steps should I take to mitigate this vulnerability?
To mitigate CVE-2026-27173, you should upgrade the Apache Airflow CNCF Kubernetes provider to version 10.17.0 or later, where the vulnerability has been fixed.
The fix involves a two-token mechanism that prevents JWT token exposure and expiration issues by issuing a longer-lived workload token with restricted scope and a short-lived execution token refreshed automatically.
Ensure that your Kubernetes environment and Airflow deployments are updated accordingly to incorporate these security improvements.
How does this vulnerability affect compliance with common standards and regulations (like GDPR, HIPAA)?:
The vulnerability exposes JWT tokens used by Kubernetes Executor workers to users with read-only access to Kubernetes Pods, potentially allowing unauthorized actions such as modifying the Airflow database state for tasks.
This unauthorized access and potential modification of task data could lead to violations of data integrity and confidentiality principles required by standards like GDPR and HIPAA, which mandate strict controls over access to sensitive data and system state.
Therefore, the vulnerability could negatively impact compliance with such regulations by enabling privilege escalation and unauthorized data manipulation.
How can this vulnerability be detected on my network or system? Can you suggest some commands?
This vulnerability involves exposure of JWT tokens used by Kubernetes Executor workers in Apache Airflow, accessible by users with read-only access to Kubernetes Pods. Detection would involve monitoring for unauthorized access or use of these JWT tokens.
Since the vulnerability relates to JWT token exposure in Kubernetes Pods, you can detect it by inspecting Kubernetes Pod logs and network traffic for suspicious token usage or unauthorized API calls to Airflow endpoints.
Suggested commands to help detect this vulnerability include:
- Use kubectl to list pods and check for unexpected access or logs: `kubectl logs <pod-name> -n <namespace>`
- Monitor Kubernetes API server audit logs for read-only users accessing pod tokens or making unauthorized API calls.
- Use network monitoring tools (e.g., tcpdump, Wireshark) to capture and analyze traffic for JWT tokens being transmitted in plaintext or suspicious API requests.
- Check Airflow task execution logs for 403 errors indicating token expiration or misuse, which may signal exploitation attempts.
Note that no specific detection commands or scripts are provided in the available resources.