CVE Database

mcp-searxng is a Model Context Protocol server that gives AI assistants web search and URL-reading capabilities through SearXNG. Prior to 1.2.0, the web_url_read URL policy in src/url-reader.ts can be bypassed while MCP_HTTP_HARDEN is enabled and MCP_HTTP_ALLOW_PRIVATE_URLS is not enabled because redirect targets are not revalidated, 0.0.0.0 is not classified as an internal address, and IPv4-mapped IPv6 literals canonicalized to hexadecimal form are not recognized. These inputs allow an attacker-influenced tool call to make the MCP server fetch loopback or internal HTTP resources and return content from local services, private APIs, service-mesh endpoints, or cloud metadata endpoints. The separate hostname-to-private-address case addressed by the earlier partial fix is not part of these residual bypasses. This issue is fixed in version 1.2.0.

mcp-searxng is a Model Context Protocol server that gives AI assistants web search and URL-reading capabilities through SearXNG. Prior to 1.2.0, web_url_read passes a caller-supplied URL to the server-side fetch path while assertUrlAllowed() in src/url-reader.ts runs only when MCP_HTTP_HARDEN is enabled, even though MCP_HTTP_HARDEN is disabled by default in src/http-security.ts. In the default configuration, an attacker who influences the URL selected by a user or AI agent can make the server fetch loopback, private-network, or cloud metadata endpoint resources and return their contents into the model context. file:// URLs remain rejected, and the separate DNS-resolution and redirect-validation bypasses are outside this record. This issue is fixed in version 1.2.0.

Meta Ads MCP is a Model Context Protocol (MCP) server that lets AI assistants run Meta Ads. Prior to version 1.0.115, the upload_ad_image tool in meta_ads_mcp/core/ads.py passes an attacker-controlled image_url to try_multiple_download_methods() in meta_ads_mcp/core/utils.py, where httpx.AsyncClient uses follow_redirects=True and performs HTTP requests without validating the scheme, host, or resolved IP address. In a streamable-http deployment, a network caller can use any non-empty authorization value because Meta credential validation occurs after the image download, then direct the server to loopback services, private-network addresses, cloud metadata endpoints, or redirect-chained internal targets. The resulting server-side request forgery can expose internal data, invoke state-changing internal services, or disrupt reachable services. This issue is fixed in version 1.0.115.

Meta Ads MCP is a Model Context Protocol (MCP) server that lets AI assistants run Meta Ads. Prior to version 1.0.115, AuthInjectionMiddleware in meta_ads_mcp/core/http_auth_integration.py rejects HTTP MCP requests only when both auth_token and pipeboard_token are absent, while extract_token_from_headers() does not recognize X-Pipeboard-Token as a primary credential. A network caller using the streamable-http transport can therefore send any X-Pipeboard-Token value, pass the guard without establishing authentication context, and cause get_auth_token() to fall back to the server operator's META_ACCESS_TOKEN. Subsequent MCP tools execute with the operator's Meta credentials and can read or modify the operator's Meta Ads data. Deployments using the default stdio transport or without META_ACCESS_TOKEN are not affected. This issue is fixed in version 1.0.115.

netty-incubator-codec-ohttp implements Oblivious HTTP (OHTTP) gateway and client functionality using Netty. Prior to 0.0.23.Final, the OHTTP gateway decryption path in codec-ohttp/src/main/java/io/netty/incubator/codec/ohttp/OHttpRequestResponseContext.java allocates a pooled direct ByteBuf for decrypted plaintext before the AEAD tag is verified. When an invalid tag causes decryptChunk() to throw CryptoException, OHttpRequestResponseContext.decodeChunk() does not release the ByteBuf because the allocation is not guarded by try/finally. Repeated invalid encrypted requests can therefore leak native off-heap memory until the gateway is unable to continue serving requests. This issue is fixed in version 0.0.23.Final.

Sakai is a Collaboration and Learning Environment (CLE). From 23.0 until 23.5 and 25.3, the DELETE /api/users/{userId}/profile/image endpoint allows an authenticated user to delete another user's profile image because ProfileController.removeProfileImage() passes the attacker-controlled userId to ProfileServiceImpl.removeProfileImage() without verifying ownership, and profileImageUploadedRepository.deleteById(userId) removes the selected row. The related DELETE /api/users/{userId}/profile/pronunciation endpoint also omits session validation and ownership checks before ProfileServiceImpl.removePronunciationRecording() deletes the target user's recording. The upload path is not affected because it already verifies ownership, and superusers remain intentionally authorized to modify other profiles. Successful exploitation can repeatedly remove profile identity artifacts, including administrator and instructor images, and disrupt workflows that rely on those artifacts. This issue is fixed in versions 23.5, 25.3, and 26.0.

Bugsink is a self-hosted error tracking tool. Prior to version 2.2.2, Bugsink stores every set of custom tags supplied with an incoming event, allowing a caller with a valid project DSN to submit an unusually large tag set and force excessive tag-row writes. Because Bugsink uses a single-writer database architecture, the expensive write transaction delays digestion of other events and causes a temporary ingestion denial of service. Version 2.2.2 applies the configurable MAX_EVENT_TAGS limit before storage. The impact is limited to availability for an instance accepting the event and does not expose stored data, modify existing events, or allow code execution. This issue is fixed in version 2.2.2.

Inspektor Gadget is a set of tools and framework for data collection and system inspection on Kubernetes clusters and Linux hosts using eBPF. From 0.27.0 until 0.53.1, the uprobe library resolver can allow an unprivileged container to consume excessive CPU and block other containers from starting by supplying a crafted /etc/ld.so.cache file while an uprobe-based gadget is active. The parser in pkg/uprobetracer/ldcache_parser.go trusts EntryCount enough to perform excessive iteration, computes cache1Len with overflowing uint32 arithmetic, and repeatedly calls readStringFromBytes in pkg/uprobetracer/bytes.go, where byte-by-byte immutable string concatenation produces quadratic work. While this processing occurs, pkg/container-hook holds the fanotify container-start pause, allowing a crafted cache to delay startup for roughly a minute, prevent Docker from starting other containers, and degrade monitoring. Processing caches from already-running containers can still consume CPU but does not hold the new-container startup pause, and the advisory establishes no confidentiality or integrity impact. This issue is fixed in version 0.53.1.

Bambuddy is a self-hosted print archive and management system for Bambu Lab 3D printers. Starting in version 0.1.6 and prior to version 0.2.4.4, a fail-open in the authentication code allows any attacker to bypass authentication by flooding a public endpoint to exhaust resources causing database access to fail, granting unauthenticated access to all protected endpoints. Version 0.2.4.4 patches the issue.

An issue in MitraStar GPT-2742GX4X5v6-SV GL_g2.5_100XNT0b23_3 allows an authenticated attacker to execute arbitrary code via the /cgi-bin/device-management-utilities-internet.cgi component

Apptainer is an open source container platform. Prior to version 1.5.1, Image.AuthorizedPath applies plain string-prefix matching to the limit container paths directive in apptainer.conf, so an allowed path such as /data/safe also authorizes a sibling path such as /data/safe-but-unsafe. A local user can consequently run a container from a directory outside the administrator's intended allowlist when Apptainer operates in setuid mode. Installations that do not use setuid mode or do not configure limit container paths are not affected, and unrestricted user namespaces already allow users to run containers of their choice. This issue is fixed in version 1.5.1.

DIRAC is an interware, meaning a software framework for distributed computing. Prior to versions 8.0.79, 9.0.22, and 9.1.10, the RequestManagementSystem/Service/ReqManagerHandler.py export_getRequestCountersWeb function passes an authenticated caller-controlled groupingAttribute to RequestManagementSystem/DB/RequestDB.py getRequestCountersWeb. An unrecognized value is resolved against the Request object and evaluated as Python code, allowing a crafted dunder attribute expression to reach operating-system functions and execute commands as the account running the DIRAC services. Successful exploitation can expose dirac.cfg, database passwords, stored proxies, and tokens, fully compromise the DIRAC system, and allow alteration of local log evidence. This issue is fixed in versions 8.0.79, 9.0.22, and 9.1.10.

OpenCost provides cost monitoring for Kubernetes workloads and cloud costs. Prior to 1.121.0, the POST /serviceKey endpoint in pkg/costmodel/router.go allows a network client to invoke AddServiceKey without mandatory authentication and submit an arbitrary key form value that is written to the GCP service-account key.json file returned by GetGCPAuthSecretFilePath in core/pkg/env/core.go. The attacker controls the file contents but not the CONFIG_PATH-derived directory, the key.json filename, or the file mode. Replacing the credential contents can disrupt GCP cost collection or cause OpenCost to use attacker-selected credentials, and the wildcard Access-Control-Allow-Origin response permits browser-assisted requests when the service is reachable from a browser. This issue is fixed in version 1.121.0.

CrowdStrike released a security update to address a vulnerability in the Falcon sensor for Windows. The vulnerability only exists when the Microsoft Office File Malicious Macro Removal Windows policy setting is enabled and customers remain protected through the Cloud Anti-malware for Microsoft Office Files settings. An update is available immediately for versions 7.34 and above, 7.32 LTS, and 7.16 for Windows 7/2008 R2 systems. The Falcon sensor for Mac, Linux, and Legacy Systems are not affected.  This vulnerability could expose an arbitrary file write to protected locations from an unprivileged context, potentially leading to local privilege escalation. The CrowdStrike Laroux Malware Cleanup Tool, based off of the same feature in the CrowdStrike Falcon sensor for Windows, is also affected. An update for this tool is also available immediately.

BharatMLStack up to and including 1.3.0 is vulnerable to Cross Site Scripting (XSS) via the component Trufflebox UI (trufflebox-ui) in ExpressionViewModal.jsx.

In BharatMLStack up to and including v1.3.0, Trufflebox UI stores the JWT authentication token, full user object, and session ID in the browser's localStorage, which is fully accessible to any JavaScript running on the page.

BharatMLStack up to and including v1.3.0 is vulnerable to Cross Site Scripting (XSS) in the component Trufflebox UI (trufflebox-ui) in GenericNumerixTable.jsx.

Concrete CMS 9.2.0 to 9.5.2 did not enforce per-field edit_user_properties permissions on the REST API user write endpoints (PUT /ccm/api/1.0/users/{uID} and POST /ccm/api/1.0/users/{uID}/change_password). A user with an update-scoped OAuth token and permission to edit only one non-sensitive field could change another non-superuser's password, username, email, and attributes, taking over that account. The Concrete CMS security team gave this vulnerability a CVSS v4.0 score of 7.4 with vector CVSS:4.0/AV:N/AC:L/AT:P/PR:H/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N. Thanks riodrwn for reporting.

In Concrete CMS 9.0 to 9.5.2, the Top Navigation Bar block did not HTML-escape dropdown child page names before writing them into the page, so a user who could create or rename pages could store a script through a child page name and have it run in the browser of any visitor, editor, or administrator who viewed the navigation and opened the affected dropdown. In the Concrete CMS origin, the script executed with the victim's privileges and could read same-origin content or perform actions available to that user. The Concrete CMS security team gave this vulnerability a CVSS v4.0 score of 7.5 with vector CVSS:4.0/AV:N/AC:L/AT:P/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N. Thanks labixiaoxin97 for reporting.

GitLab has remediated an issue in GitLab CE/EE affecting all versions from 15.7 before 19.1.8, 19.2 before 19.2.6, and 19.3 before 19.3.2 that under certain conditions could have allowed an authenticated user to access CI/CD variables outside their intended environment scope due to improper input validation in the environment scope pattern matcher.

GitLab has remediated an issue in GitLab CE/EE affecting all versions from 18.6 before 19.1.8, 19.2 before 19.2.6, and 19.3 before 19.3.2 that under certain conditions could have allowed an authenticated user to bypass SAML SSO sign-in restrictions and authenticate without SSO due to missing authentication enforcement checks.

IBM Business Automation Workflow containers and traditional is vulnerable to an XML external entity injection (XXE) attack when processing XML data. A remote attacker could exploit this vulnerability to expose sensitive information or consume memory resource.

IBM Cloud Pak for Business Automation is vulnerable to HTML injection. A remote attacker could inject malicious HTML code, which when viewed, would be executed in the victim's Web browser within the security context of the hosting site.

IBM Cloud Pak for Business Automation is vulnerable to stored cross-site scripting. This vulnerability allows an authenticated user to embed arbitrary JavaScript code in the Web UI thus altering the intended functionality potentially leading to credentials disclosure within a trusted session.

IBM Cloud Pak for Business Automation is vulnerable to stored cross-site scripting. This vulnerability allows an authenticated user to embed arbitrary JavaScript code in the Web UI thus altering the intended functionality potentially leading to credentials disclosure within a trusted session.

IBM Business Automation Workflow containers and traditional could allow an authenticated attacker to trigger restricted import actions due to missing authorization controls.

IBM MQ 9.1.0.0 through 9.1.0.37 LTS, 9.2.0.0 through 9.2.0.43 LTS, 9.3.0.0 through 9.3.0.41 LTS, 9.3.0.0 through 9.3.5.1 CD, 9.4.0.0 through 9.4.0.25 LTS, 9.4.0.0 through 9.4.5.1 CD, and 10.0.0.0 could allow an authenticated attacker to execute arbitrary code due to a deserialization of untrusted data.

IBM MQ 9.1.0.0 through 9.1.0.37 LTS, 9.2.0.0 through 9.2.0.43 LTS, 9.3.0.0 through 9.3.0.41 LTS, 9.3.0.0 through 9.3.5.1 CD, 9.4.0.0 through 9.4.0.25 LTS, 9.4.0.0 through 9.4.5.1 CD, and 10.0.0.0 could allow an authenticated attacker to read files from a vulnerable .NET client or cause limited denial of service due to improper handling of XML external entities in RFH2 folder parsing.

IBM MQ 9.1.0.0 through 9.1.0.37 LTS, 9.2.0.0 through 9.2.0.43 LTS, 9.3.0.0 through 9.3.0.41 LTS, 9.3.0.0 through 9.3.5.1 CD, 9.4.0.0 through 9.4.0.25 LTS, 9.4.0.0 through 9.4.5.1 CD, and 10.0.0.0 Classes for Java could allow an authenticated attacker to obtain sensitive information or cause a denial of service due to XML external entity injection in MQRFH2 header processing.

IBM Verify Identity Access could allow a remote attacker to cause a denial of service due to insufficient validation of incoming request resources.