CVE Vulnerability Catalog

The Azure Key Vault secrets backend in Apache Airflow's Microsoft Azure provider resolved a team-scoped Connection or Variable id through the team-agnostic lookup when the team-scoped lookup missed. In a deployment running multi-team mode with this backend, a caller in one team could resolve a secret belonging to another team by supplying an id that spells out that team's namespace, obtaining its credentials in full. Users are advised to upgrade to apache-airflow-providers-microsoft-azure 14.1.0 or later, which refuses the team-agnostic fall-through for an id that could name a team namespace.

A Zip Slip vulnerability in the WebUI ISP Upgrade functionality allows arbitrary file write via a crafted archive containing directory traversal sequences. An authenticated administrator may overwrite arbitrary files on the system. Successful exploitation may allow arbitrary file to be overwritten on the underlying system, affecting system integrity and availability.

Metabase allows an authenticated, low-privileged attacker to read the entire Metabase application database.

Dokploy prior to 0.29.13 contains a vulnerability in the createCommand() function that constructs shell commands by interpolating compose service names and configuration into bash command strings. When a compose with a maliciously crafted name or service definition is deployed, shell metacharacters are interpreted as command separators, allowing arbitrary command execution on the Docker host.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 contains a vulnerability in the discourse_templates endpoint that exposed hidden tag names. The DiscourseTemplates::TemplatesSerializer did not filter tags through the request Guardian, so it did not respect tag group permissions, allowing users to see tags they were not permitted to view.

tar-rs versions 0.4.11 through 0.4.46 contain a symlink escape vulnerability in the Builder::append_dir_all() function that allows attackers to read files outside the intended source root directory by planting symlinks in an attacker-controlled directory. When a privileged process archives an untrusted directory, the function follows symlinks without verifying that resolved targets remain within the source root, causing out-of-bounds files to be included in the archive as regular files and disclosed to the attacker.

Discourse, an open-source discussion platform, prior to 2026.1.7 allowed an authenticated user to submit specially formed URLs that bypassed the Onebox allowlist and embedded malicious content in a site. This issue is fixed in versions 2026.1.7, 2026.6.2, 2026.7.1, and 2026.8.0-latest.1.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 allowed a low-privileged user to place crafted content in the moderation review queue that executed stored cross-site scripting when a moderator viewed it on a site with a modified or disabled default Content Security Policy. This issue is fixed in versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0.

In multicluster-global-hub, during a ManagedClusterMigration, the system incorrectly grants all managed hubs read access to a shared communication topic. This allows a compromised managed hub to intercept and collect sensitive bootstrap kubeconfigs, which contain API server tokens intended for other hubs. These tokens have an extended validity of approximately 9.86 years, significantly increasing the risk of unauthorized access and information disclosure to other managed clusters.

In libvirt, during storage volume clone or convert operations, newly created volume images were temporarily world-readable. This was caused by the qemu-img utility running with overly permissive file creation settings, allowing any local user to read the full guest disk contents. This vulnerability could lead to sensitive information disclosure from guest virtual machines.

HCL BigFix Mobile is vulnerable to Reflected Cross-Site Scripting (Reflected XSS) due to insufficient validation and output encoding of user-controlled input.

Vault's ACL policy engine did not consistently enforce a wildcard (glob) deny rule against LIST requests made with a trailing slash on the denied path. A token with a broader allow rule and a narrower wildcard deny rule may enumerate entry names beneath a path it was intended to be denied access to.

Discourse, an open-source discussion platform, prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 allowed an authenticated user to eavesdrop on private AI bot conversations through the AI bot reply stream.

Discourse prior to version 2026.1.6 rendered previous and new value fields in the staff action log model without escaping, which could inject stored cross-site scripting into the staff interface.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 in plugins/chat/lib/chat/onebox_handler.rb resolves Chat::Thread by route thread_id independently of the route channel_id before checking whether the user can preview the selected chat channel. An authenticated user can pair a public channel ID with a private thread ID in a /onebox.json request and obtain private thread message content.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 in SiteSerializer.anonymous_default_navigation_menu_tags serializes tags from SiteSetting.default_navigation_menu_tags without applying DiscourseTagging.filter_visible for the anonymous viewer. An unauthenticated user can retrieve restricted tag names and descriptions through /site.json when those tags are limited by inaccessible categories, category tag groups, or tag-group permissions.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 in TopicLink.extract_from, TopicLink.ensure_entry_for, and TopicLink.duplicate_lookup does not consistently enforce Guardian.can_see? checks when processing internal links. An authenticated user can submit links to restricted topics, private messages, or hidden posts and receive canonicalized slugs or titles in the composer_messages duplicate_lookup response even though the targets are not visible to that user.

Discourse prior to versions 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 in Onebox::DomainChecker.is_blocked? compares hostnames and SiteSetting.blocked_onebox_domains entries case-sensitively, allowing an attacker to bypass configured Onebox domain restrictions by changing character casing in a redirect target hostname.

Discourse prior to versions 2026.1.7, 2026.6.2, 2026.7.1, and 2026.8.0-latest.1 has HTML injection in PrettyText.format_for_email because cooked attribute values are reparsed as markup. Crafted Vimeo iframe sources, secure-upload URLs or dimensions, and hashtag data-slug values can cause decoded attribute text to be reinterpreted as HTML. The vulnerable conversion also fails to strictly validate the Vimeo iframe host and path, allowing non-Vimeo allowlisted iframes to be converted.

Chatwoot prior to version 4.9.0 allowed authenticated account administrators to transfer Portals, Automation Rules, Macros, and Twilio Channels to other accounts through the writable account_id parameter. This could break tenant isolation and cause cross-account data exposure, unauthorized configuration changes, or loss of access to transferred resources.