CVE-2025-0453: Denial of Service through Batched Queries in GraphQL in mlflow/mlflow
In mlflow/mlflow version 2.17.2, the /graphql endpoint is vulnerable to a denial of service attack. An attacker can create large batches of queries that repeatedly request all runs from a given experiment. This can tie up all the workers allocated by MLFlow, rendering the application unable to respond to other requests. This vulnerability is due to uncontrolled resource consumption.
Affected Software
Event History
Frequently Asked Questions
What is the severity of CVE-2025-0453?
CVE-2025-0453 has a high severity due to its potential for a denial of service attack.
How do I fix CVE-2025-0453?
To fix CVE-2025-0453, upgrade to a patched version of MLflow that addresses this vulnerability.
What systems are affected by CVE-2025-0453?
CVE-2025-0453 affects MLflow version 2.17.2 and potentially earlier versions.
What attack vector does CVE-2025-0453 utilize?
CVE-2025-0453 allows attackers to exploit the `/graphql` endpoint by submitting large batches of queries.
What impact does CVE-2025-0453 have on MLflow?
CVE-2025-0453 can tie up all workers within MLflow, leading to service unavailability.