Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessAuthenticate as low/no-role AAP user
-
DeliveryQuery ancestor_artifacts via REST filter/regex lookup
-
ExploitUse result count as boolean oracle
-
ExecutionIterate character-by-character across positions
-
ImpactRecover no_log secrets cross-organization
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V8 - Authorization Object-level authorization, IDOR prevention
|
| Exploitation Conditions | Requires a valid authenticated account on the automation-controller: either read access to a workflow (PR:L path), or - via a regex lookup that bypasses the JSON cross-relation filter guard through the world-readable credential-types endpoint - any authenticated user with no roles at all. The target secrets must have been propagated as merged set_stats artifacts into WorkflowJobNode.ancestor_artifacts and persisted before Ansible's no_log masking, so exploitation depends on workflows that use set_stats to pass data between nodes. Limiting factors: an account is mandatory (not internet-anonymous), and recovery is an incremental count/boolean oracle requiring many requests per secret character, making it slow and noisy rather than a single-shot read. |
| Impact | This is a genuine but bounded confidentiality issue, not a top-tier emergency. The Red Hat CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, base 6.5) correctly captures a network-exploitable, low-complexity flaw needing only a low-privileged authenticated account, with pure confidentiality impact (C:H, no integrity or availability effect) - consistent with CWE-639 (authorization bypass through user-controlled key / IDOR). No EPSS score, no CISA KEV listing, and no public POC were provided, so exploitation status is 'no public exploit identified at time of analysis'; this argues against imminent mass exploitation. The severity is driven by impact, not ease of mass-abuse: it is a slow character-by-character oracle attack requiring many crafted REST filter requests, so it favors a targeted, patient adversary already inside the platform over opportunistic scanning. One conflict to verify: the vector states PR:L, but the description notes a second path via the world-readable credential-types endpoint usable by 'any authenticated user with no roles,' which is functionally near-unauthenticated-insider access - the effective barrier may be lower than PR:L implies, though a valid account is still required. Cross-org secret recovery (no_log-masked values) meaningfully raises the stakes in multi-tenant AAP deployments. |
| Exploit ScenarioDerived | A low-privileged authenticated AAP user (in the worst case, a valid account with no assigned roles) issues crafted REST API filter queries against the WorkflowJobNode.ancestor_artifacts column using regular-expression/count lookups, reading the result count as a boolean oracle. By iterating character positions, they reconstruct secret values a playbook author marked no_log - including secrets belonging to other organizations. No public exploit code is identified, but the technique is deterministic and scriptable given the described filter behavior. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Sep 23, 2026
Secret values that playbook authors explicitly marked with no_log can be recovered from Red Hat Ansible Automation Platform 2 (automation-controller) through the REST filter backend, because the WorkflowJobNode.ancestor_artifacts column is not wrapped in prevent_search() and therefore remains queryable even though it is omitted from the API serializer. An authenticated user - one with only read access to a workflow, or any authenticated user with no roles at all by abusing a regular-expression lookup that bypasses the JSON cross-relation filter guard via the world-readable credential-types endpoint - can use the result count as a boolean/count oracle to extract secrets character by character, potentially across organizations. The issue is rated CVSS 6.5 (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, CWE-639); no public exploit code or confirmed active exploitation was identified at time of analysis, and because recovery requires an account plus many incremental requests per character, the practical risk is a bounded confidentiality exposure rather than a top-tier emergency.
RemediationAI
No vendor-released patch is identified in the available data for CVE-2026-84720; the authoritative fix status must be confirmed against the Red Hat advisory at https://access.redhat.com/security/cve/CVE-2026-84720 and the Bugzilla record at https://bugzilla.redhat.com/show_bug.cgi?id=2527214, and administrators should upgrade to the first automation-controller build Red Hat marks as fixed once it is published. Until then, apply compensating controls that directly address the exposed lookup path: at a reverse proxy or WAF in front of the controller API, block request parameters containing regex/count lookup suffixes (for example __regex, __iregex, and __count) against workflow-job-node and credential-type endpoints - the trade-off is that legitimate advanced filtering used by integrations or scripts will break, so scope the rule narrowly and log matches; and restrict access to the credential-types endpoint to trusted roles if it is not required by ordinary users, though this does not fully close the workflow-read-access path. Because the permissions model is the primary throttle, audit and minimize who holds workflow read or any authenticated roles, and rotate any secrets that were propagated via set_stats across workflow nodes, since the persisted ancestor_artifacts values may already have been partially extracted. Detection should focus on anomalous, high-volume API traffic exhibiting many repeated filter requests with differing regex patterns or count operations against the same endpoint - a signature consistent with the character-by-character oracle - and on removal of secrets from set_stats-based inter-node data passing in favor of credential lookups, accepting the operational redesign cost that entails. Note that exploitation requires a valid authenticated account and produces slow, noisy enumeration rather than a single-shot disclosure, so these measures reduce an already bounded exposure.
Affected VersionsENISA
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
The root cause is CWE-639 (Authorization Bypass Through User-Controlled Key): a data field that the API intentionally hides from serialized output is still reachable as a filterable field on the queryset. In AWX/automation-controller (Django REST Framework based), fields that must not be exposed to client-controlled lookups are normally protected by the prevent_search() decorator, which removes them from the search/filter backend. Here, WorkflowJobNode.ancestor_artifacts - the column holding the raw merged set_stats artifacts that workflows propagate between nodes - was never wrapped, so clients can append lookups such as regular expressions or counts against it in API filter parameters. Two properties compound the problem: the column is persisted before Ansible applies its no_log masking, meaning the stored value is the unmasked secret; and because the column is not serialized, an attacker cannot simply read it, but can still observe aggregate results (record counts, boolean truth of a match) to build a character-by-character extraction oracle. The reported bypass additionally routes a regex lookup through the world-readable credential-types endpoint, sidestepping the JSON cross-relation filter guard, which lowers the required privilege from workflow read access to any authenticated account. Two different CPE strings are listed for the same product family (Red Hat Ansible Automation Platform 2) with wildcard versions, indicating that all shipped 2.x builds of the automation-controller are considered in scope until Red Hat publishes version-specific fixed builds.
Know when Django gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
Network REST API, low complexity, requires a valid (even role-less) account so PR:L; pure confidentiality disclosure of secrets, no integrity or availability impact.
Primary rating from Vendor (redhat).
CVSS VectorVendor: redhat
DescriptionCVE.org
A flaw was found in the Ansible Automation Platform automation-controller. The WorkflowJobNode.ancestor_artifacts database column, which stores the raw merged set_stats artifacts propagated between workflow nodes, is not wrapped in prevent_search() and is therefore accepted for arbitrary field lookups by the REST filter backend, even though it is omitted from the API serializer. Because the column is persisted before Ansible's no_log masking is applied, a user with only read access to a workflow - or, via a regular-expression lookup that bypasses the JSON cross-relation filter guard through the world-readable credential-types endpoint, any authenticated user with no roles - can use the result count as a boolean/count oracle to recover, character by character, secret values that a playbook author explicitly marked no_log, including across organizations.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
The package python/cpython from 0 and before 3.6.13, from 3.7.0 and before 3.7.10, from 3.8.0 and before 3.8.8, from 3.9
In Django 1.10.x before 1.10.8 and 1.11.x before 1.11.5, HTML autoescaping was disabled in a portion of the template for
An issue was discovered in Django 3.2 before 3.2.14 and 4.0 before 4.0.6. Rated critical severity (CVSS 9.8), this vulne
Django 3.1.x before 3.1.13 and 3.2.x before 3.2.5 allows QuerySet.order_by SQL injection if order_by is untrusted input
Django before 1.11.27, 2.x before 2.2.9, and 3.x before 3.0.1 allows account takeover. Rated critical severity (CVSS 9.8
An authentication/authorization bypass in the djust Python live-view framework (versions prior to 1.0.7) lets unauthenti
Unauthenticated arbitrary Python module import in djust before 1.0.7 lets any client that can reach the WebSocket or SSE
Django 1.11 before 1.11.29, 2.2 before 2.2.11, and 3.0 before 3.0.4 allows SQL Injection if untrusted data is used as a
Cross-site scripting (XSS) vulnerability in the dismissChangeRelatedObjectPopup function in contrib/admin/static/admin/j
Forged content modification in django-page-cms through 2.0.13 lets unauthenticated remote attackers exploit five CSRF-ex
Cross-site request forgery in djust's Server-Sent-Events transport affects all versions prior to 1.0.7 and allows a cros
Unauthenticated remote attackers can read live Django view and session state and invoke server-side handlers through dju
EUVD-2026-85614
GHSA-h789-rmcq-fh7h
NVD
CVE.org
EPSS
Vulnrichment
Exploit-DB