What is the CVSS score of CVE-2026-45360?

CVE-2026-45360 has a CVSS 3.1 base score of 7.3 (High). CVSS vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L. EPSS exploitation probability: 0.0%.

Which versions of Apache Airflow are affected by CVE-2026-45360?

Apache Airflow versions prior to 3.2.2 are vulnerable to arbitrary code execution through malicious serialized objects processed by the scheduler, potentially compromising data pipeline integrity and…. A patch is available.

How to fix or mitigate CVE-2026-45360?

Within 24 hours: Identify and document all Apache Airflow instances and current versions in use. Within 7 days: Upgrade all instances to Apache Airflow 3.2.2 or later, starting with production schedulers. Within 30 days: Verify patches are applied across all environments and audit for any suspicious scheduler logs or unexpected code execution during the vulnerable period.

Apache Airflow CVE-2026-45360

EUVD-2026-33587 HIGH

Deserialization of Untrusted Data (CWE-502)

2026-05-31

GHSA-2r5m-76wx-56gx

Apache Deserialization Python Airflow

7.3

CVSS 3.1 · Vendor

Severity by source

Vendor (CNA) PRIMARY

7.3 HIGH

AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L

Primary rating from Vendor (CNA) · only source for this CVE.

CVSS VectorVendor

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L

Attack Vector

Network

Attack Complexity

Low

Privileges Required

None

User Interaction

None

Scope

Unchanged

Confidentiality

Low

Integrity

Low

Availability

Low

Lifecycle Timeline

Source Code Evidence Fetched

Jun 02, 2026 - 17:24 vuln.today

Analysis Generated

Jun 02, 2026 - 17:24 vuln.today

CVSS changed

Jun 02, 2026 - 17:22 NVD

7.3 (HIGH)

Patch available

Jun 01, 2026 - 10:01 EUVD

CVE Published

May 31, 2026 - 12:45 nvd

HIGH 7.3

CVE Published

May 31, 2026 - 12:45 nvd

UNKNOWN (no severity yet)

Blast Radius

ecosystem impact

† from your stack dependencies † transitive graph · vuln.today resolves 4-path depth

8 pypi packages depend on apache-airflow (4 direct, 4 indirect)

Ecosystem-wide dependent count for version 3.2.2.

Description PRE-NVD

Disclosed via oss-security. NVD scoring and full description are pending.

AnalysisAI

Arbitrary Python module import in Apache Airflow versions prior to 3.2.2 occurs when the scheduler deserializes custom DeadlineReference objects, because the prior implementation called import_string() directly on an attacker-controllable __class_path field. Rated CVSS 7.3 with low confidentiality/integrity/availability impact, this issue has no public exploit identified at time of analysis and EPSS estimates exploitation probability at 0.02% (6th percentile).

Technical ContextAI

Apache Airflow's serialization layer (airflow/serialization/definitions/deadline.py) reconstructs custom DeadlineReference subclasses from serialized DAG/deadline metadata. Before the fix, deserialize_reference() called airflow._shared.module_loading.import_string(reference_data['__class_path']), which performs a Python import of any dotted module path supplied in the serialized payload. This is a classic CWE-502 (Deserialization of Untrusted Data) weakness: importing a module executes its top-level code, so an attacker who can influence the class_path string can trigger side effects of arbitrary importable modules within the scheduler process - including modules whose import has destructive side effects, opens network connections, or otherwise affects Python state. The patch (PR #66737) replaces the unrestricted import_string call with find_registered_custom_deadline_reference(), which only resolves classes that have been explicitly enrolled via a new AirflowPlugin.deadline_references registry, matching the existing pattern Airflow uses for custom timetables and partition mappers.

RemediationAI

Upgrade Apache Airflow to version 3.2.2 or later, which replaces the unrestricted import_string() call in deadline deserialization with a plugin-registry lookup (see PR https://github.com/apache/airflow/pull/66737 and the Apache announcement at https://lists.apache.org/thread/q227dghjwgfz8xsxrf2pwpz4wk43zm83). After upgrading, any custom DeadlineReference subclasses must be explicitly registered via the new deadline_references attribute on an AirflowPlugin, or scheduler-side deserialization will raise DeadlineReferenceNotRegistered - operators using custom deadline references should plan a small refactor to add the plugin entry. Where immediate upgrade is not possible, restrict who can author DAGs and who can write to the Airflow metadata database to trusted users only, audit existing serialized deadline payloads for unexpected __class_path values, and consider running schedulers in a sandboxed Python environment to limit the side effects of arbitrary imports; these workarounds reduce exposure but do not close the import primitive itself.