CVE-2026-48129: Kestra Path Traversal in inputFiles – Patch Guidance

Technical summary

The vulnerability resides in how Kestra's task engine handles the `inputFiles` parameter. Prior to the patched versions (1.3.19, 1.2.19, 1.1.19, 1.0.43), the platform fails to sanitize file name inputs before resolving them on disk. When untrusted data—such as webhook payloads or forwarded execution context—flows into file name fields without validation, an attacker can craft payloads containing path traversal sequences. The rendered file names are then written directly to the filesystem, allowing traversal outside the task's working directory. This is a classic CWE-22 path traversal issue amplified by the trust boundary between user-controlled input and file I/O operations.

Business impact

Exploitation could lead to unauthorized file creation or modification on worker nodes, potentially compromising application integrity and confidentiality. In multi-tenant or shared orchestration environments, an attacker exploiting a vulnerable flow could overwrite configuration files, inject malicious code into application directories, or corrupt data relied upon by other workflows. The blast radius depends on worker process privileges and network isolation; however, even in restricted deployments, the ability to place files in unexpected locations poses operational risk.

Affected systems

Kestra versions prior to 1.3.19, 1.2.19, 1.1.19, and 1.0.43 are vulnerable. Organizations running any older version within the 1.0, 1.1, 1.2, or 1.3 release families should prioritize identifying and upgrading affected instances. The vulnerability applies to self-hosted and container-deployed Kestra environments where flows process external input.

Exploitability

Exploitation requires the attacker to influence data that flows into an `inputFiles` field—either by controlling a webhook source, crafting a pull request that modifies a flow definition, or compromising an upstream service that feeds execution data to Kestra. The CVSS score of 6.5 reflects medium risk: network-accessible but requiring a somewhat specialized attack chain (AC:H) and no user interaction. Real-world impact depends on whether flows are designed to ingest untrusted data and whether deployed workers have write access to critical system directories.

Remediation

Upgrade affected Kestra instances to version 1.3.19, 1.2.19, 1.1.19, or 1.0.43 or later. Organizations should also review existing flows to identify those that accept external or webhook-driven input and apply them to `inputFiles` parameters, enforcing additional validation on file names where feasible. In the interim, restrict flow definitions to trusted sources and audit worker-level file system permissions to limit the impact of unauthorized writes.

Patch guidance

Apply the patched versions according to your Kestra deployment model: for containerized deployments, update the image tag; for binary/compiled installs, download and deploy the new release. Verify the upgrade by checking the version output (e.g., `kestra --version` or via the web UI). Test a non-critical flow that uses `inputFiles` to confirm functionality before rolling out across production workers. Stagger updates if running a distributed worker pool to maintain orchestration availability.

Detection guidance

Monitor worker filesystem logs and audit trails for unexpected file creation or modification outside standard task working directories. Search for flow definitions that reference `inputFiles` with dynamic, untrusted inputs—especially those accepting webhook data or variable substitution from external sources. Review task execution logs for unusual path segments (`../`, `..\`) in resolved file names. On the host level, monitor for file writes to sensitive directories (e.g., `/etc`, `/var`, application config folders) from Kestra worker processes.

Why prioritize this

Although not yet in CISA's KEV catalog, this vulnerability merits prioritization because it enables arbitrary file writes in multi-tenant and externally-facing orchestration environments. The combination of network accessibility and potential for lateral movement (via corrupted config or injected code) justifies treating it as a near-term patch candidate, particularly for organizations that operate Kestra in shared or cloud-native environments.

Risk score, explained

The CVSS 6.5 (Medium) score reflects: (1) network-accessible attack vector, no authentication required; (2) attack complexity marked high, indicating a non-trivial chain (flow design, data flow, path traversal payload construction); (3) impact limited to integrity (file writes) with some availability risk if critical files are overwritten. The score does not account for context-specific factors such as worker isolation, multi-tenancy, or regulatory compliance—organizations should adjust internal risk ratings based on their deployment posture.

Frequently asked questions