Django Patches Four Critical Vulnerabilities, Including Potential RCE
Django has released urgent security updates for versions 6.0.8 and 5.2.17, addressing four vulnerabilities, one of which could lead to remote code execution.
The Django development team has issued critical security updates for its popular Python web framework, releasing Django 6.0.8 and Django 5.2.17 to patch a total of four vulnerabilities. Developers and system administrators are strongly advised to upgrade their deployments immediately, particularly those utilizing Django's spatial features or exposing the built-in admin interface to staff users.
The most severe of the patched flaws, identified as CVE-2026-15307, is rated high severity. This vulnerability resides within Django's spatial lookups, specifically in how it processes raster-related values through the GDALRaster library. Previously, the framework could incorrectly accept string and dictionary values when they were intended to represent rasters. In certain configurations, depending on the specific raster driver in use, a maliciously crafted input could trigger an unintended file write operation on the server or initiate a network request, all executed with the permissions of the Django process. In environments where staff users with view permissions could filter models containing spatial fields via the Django admin changelists, this could potentially pave the way for remote code execution.
To mitigate CVE-2026-15307, Django now actively blocks dictionary values and strings that do not conform to valid GEOSGeometry object formats from being used in spatial lookups. While this change is crucial for security, it may impact applications that previously relied on these less-strict input types. Django has clarified that direct model field assignments remain unaffected and will continue to accept these values. The framework recommends that developers handling untrusted spatial data should implement robust validation before use and consult Django's specific raster security guidance if they need to restore similar functionality within custom code.
Two denial-of-service (DoS) vulnerabilities were also addressed. CVE-2026-15337, a low-severity flaw in django.utils.translation.check_for_language(), could allow an attacker to consume excessive server memory by repeatedly sending long language codes via POST requests to the optional set_language() view. While Django's request-size settings and a maximum cache-entry count offered some protection, the update now rejects language codes exceeding 500 characters before they are cached, significantly reducing memory footprint.
The second DoS vulnerability, CVE-2026-15830, affects applications performing geographic geometry processing. Deeply nested GEOMETRYCOLLECTION objects, when processed by the underlying GEOS library, could lead to a segmentation fault, potentially crashing the Django application. This could be exploited by an attacker who crafts specific geometry input passed to spatial field lookups or a GeometryField form field. The fix enforces a maximum of 198 nested geometry collections for both well-known text (WKT) and well-known binary (WKB) inputs. Users requiring a different threshold can configure this via the new max_geom_collections setting in GEOSGeometry, as well as in relevant form and model fields. GeoJSON inputs are not affected as they are parsed by GDAL.
Finally, CVE-2026-15920 addresses a moderate-severity stored cross-site scripting (XSS) risk within Django's admin views. The URLField values were previously rendered as clickable links in changelists and read-only admin fields without proper validation of the URL scheme. This allowed for the rendering of potentially dangerous schemes as active links. Django now employs the URLValidator to check URLField values before rendering them as links, displaying invalid values as plain text instead.
These security fixes have also been integrated into Django's main development branch and the Django 6.1 release candidate. Organizations running Django are urged to prioritize these updates, test the compatibility of their spatial lookup implementations, and review their access control and input validation practices for the admin interface.
The new article details two additional critical vulnerabilities affecting Veeam Service Provider Console (CVE-2026-58073, CVSS 9.5) and HashiCorp's Terraform MCP Server (CVE-2026-16498, CVSS 10.0). The Veeam flaw allows unauthenticated attackers to obtain managed agent credentials, while the HashiCorp vulnerability enables cross-tenant token reuse in stateless HTTP mode. Both vendors have released patches for these issues, alongside the Django vulnerabilities previously reported.