Vulnerabilities (CVE)

Seven Gitea API routes under `/api/v1/users/{username}` (followers, following, following/{target}, starred, subscriptions, keys and gpg_keys) were registered without the `individualPermsChecker` middleware that enforces the visibility of the target user. An authenticated user could therefore read the followers, following list, starred and watched repositories and public SSH and GPG keys of a user with limited or private visibility, whom they could not otherwise see. The issue was fixed in 1.27.3 and 28.0.0 and disclosed with 28.2.0.Show less

Deleting a label compared the ID of the repository in the request against the organization ID of organization labels, and the two ID sequences overlap. A user with write access to the issues of a repository could therefore delete the labels of an organization that has the same numeric ID as that repository. Labels are now deleted only within their own repository or organization.Show less

Adding attachments to a release checked the repository and release of an attachment but not whether it was already linked to an issue or comment. A user with write access to the releases of a repository, who knew the UUID of an attachment in that repository, could link it to a release. This exposed the file name, size and download count of the attachment through the release and allowed the user to rename or delete it.Show less

The visibility check of Gitea user profiles was applied to GET requests only, while `POST /{username}` (follow and unfollow) required only a signed-in user. Any authenticated user could therefore obtain the full name, location, website and description of a user with a private profile from the response. The issue was fixed in 28.1.0 and disclosed with 28.2.0.Show less

Creating a draft release through the API returned an `assets` list that was built from a release without an ID, which matched every attachment that was not linked to a release. A user who could create a draft release in a repository therefore received the metadata (file name, UUID, upload time, size and download count) of issue and comment attachments of repositories they could not read, and of abandoned uploads.Show less

`GET /api/v1/user/subscriptions` returned the metadata of every repository a user was watching without checking that the user still had access to it. A user who watched a private repository while having access, and later lost it, for example through a team permission downgrade, could still read its name, description, clone URLs, owner and counters. The issue was fixed in 1.27.0 and 28.0.0 and disclosed with 28.2.0.Show less

Gitea resolved the principals of an SSH certificate by searching the content of all stored public keys instead of only principal keys. When Gitea trusts a user certificate authority (`SSH_TRUSTED_USER_CA_KEYS`) and accepts arbitrary principals (`SSH_AUTHORIZED_PRINCIPALS_ALLOW=anything`), a user could register another user's public key line as their own principal, or a holder of a certificate signed by that authority could list it, and then authenticate as the owner of that key through the HTTP signature API and the built-in SSH server. Instances that do not trust a user certificate authority are not affected.Show less

The Gitea web handlers for adding and removing reactions on issues and comments did not refuse locked issues for users without write access, although the API did. A signed-in user who could read an issue could therefore add or remove reactions on a locked issue and its comments, bypassing the interaction restriction of the lock. Only reactions were affected; no content could be changed.Show less

`POST /{owner}/{repo}/topics` was registered without the sign-in middleware that provides the cross-origin protection, so a request issued by a third-party site was not rejected. A repository administrator who visited a malicious page could have the topics of their repository changed. No other repository settings were reachable through this route.Show less

The Gitea Actions artifact v4 (Twirp) handlers read the whole request body into memory without a size limit. A user who can run a workflow and obtains the runner token of that run could send an oversized request to the artifact service and exhaust the memory of the server, which terminates the Gitea process. The request body is now limited to 1 MiB.Show less

Requesting `/{username}/-/repositories` for an individual user instead of an organization made the organization home handler dereference a nil organization and panic. The request needs no authentication and no special configuration. The server answered with an HTTP 500 error and logged a panic for each request; the fix returns 404 for individual users.Show less

The Gitea OpenID association callback in the security settings accepted a positive OpenID 2.0 assertion without checking that the signed-in user had started the linking. Because OpenID 2.0 allows unsolicited assertions, an attacker-controlled OpenID provider could attach the attacker's identity to the account of a signed-in victim who opened a crafted link, and the attacker could then sign in as the victim through OpenID login, which does not apply the local TOTP or WebAuthn checks. The callback now requires a time-limited `state` value bound to the user.Show less

JQ::Lite versions before 2.51 for Perl execute Perl code embedded in a query's regular expression by compiling it with string eval. The pattern given to test(), match(), or the match and =~ conditions of select() is placed in a `qr'...'` literal and compiled as source text, so a `(?{ ... })` code block in it runs without `use re 'eval'`. Single quotes in the pattern are escaped, but a code block runs in place and needs none. An attacker who supplies a query, or text that an application inserts into the pattern of one, can execute arbitrary Perl code in the calling process.Show less

The Management Console fails to properly validate input related to the configuration of JDBC user stores. This weakness allows an attacker to supply specially crafted input that can lead to the execution of arbitrary code on the affected server. Successful exploitation of this vulnerability enables an attacker to execute arbitrary code on the server with the privileges of the WSO2 product. This could lead to a full compromise of the system.Show less

Cal.com is open-source scheduling software. Prior to 5.9.8, A flaw in the login credentials provider allows an attacker to bypass password verification when a TOTP code is provided, potentially gaining unauthorized access to user accounts. This issue exists due to problematic conditional logic in the authentication flow. This vulnerability is fixed in 5.9.8.Show less

Vim is an open source, command line text editor. Prior to version 9.1.1947, an uncontrolled search path vulnerability on Windows allows Vim to execute malicious executables placed in the current working directory for the current edited file. On Windows, when using cmd.exe as a shell, Vim resolves external commands by searching the current working directory before system paths. When Vim invokes tools such as findstr for :grep, external commands or filters via :!, or compiler/:make commands, it may inadvertently run a malicious executable present in the same directory as the file being edited. The issue affects Vim for Windows prior to version 9.1.1947.Show less

The Aimeos GrapesJS CMS extension provides page editor for creating content pages based on extensible components. Prior to 2021.10.8, 2022.10.8, 2023.10.8, 2024.10.8, and 2025.10.8, Javascript code can be injected by malicious editors for a stored XSS attack if the standard Content Security Policy is disabled. This vulnerability is fixed in 2021.10.8, 2022.10.8, 2023.10.8, 2024.10.8, and 2025.10.8.Show less

Arcade MCP allows you to to create, deploy, and share MCP Servers. Prior to 1.5.4, the arcade-mcp HTTP server uses a hardcoded default worker secret ("dev") that is never validated or overridden during normal server startup. As a result, any unauthenticated attacker who knows this default key can forge valid JWTs and fully bypass the FastAPI authentication layer. This grants remote access to all worker endpoints—including tool enumeration and tool invocation—without credentials. This vulnerability is fixed in 1.5.4.Show less

vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.11.1, vllm has a critical remote code execution vector in a config class named Nemotron_Nano_VL_Config. When vllm loads a model config that contains an auto_map entry, the config class resolves that mapping with get_class_from_dynamic_module(...) and immediately instantiates the returned class. This fetches and executes Python from the remote repository referenced in the auto_map string. Crucially, this happens even when the caller explicitly sets trust_remote_code=False in vllm.transformers_utils.config.get_config. In practice, an attacker can publish a benign-looking frontend repo whose config.json points via auto_map to a separate malicious backend repo; loading the frontend will silently run the backend’s code on the victim host. This vulnerability is fixed in 0.11.1.Show less

HTCondor Access Point before 25.3.1 allows an authenticated user to impersonate other users on the local machine by submitting a batch job. This is fixed in 24.12.14, 25.0.3, and 25.3.1. The earliest affected version is 24.7.3.Show less

WebPros Plesk before 18.0.73.5 and 18.0.74 before 18.0.74.2 on Linux allows remote authenticated users to execute arbitrary code as root via domain creation. The attacker needs "Create and manage sites" with "Domains management" and "Subdomains management."Show less

The MCP Python SDK, called `mcp` on PyPI, is a Python implementation of the Model Context Protocol (MCP). Prior to version 1.23.0, tThe Model Context Protocol (MCP) Python SDK does not enable DNS rebinding protection by default for HTTP-based servers. When an HTTP-based MCP server is run on localhost without authentication using FastMCP with streamable HTTP or SSE transport, and has not configured TransportSecuritySettings, a malicious website could exploit DNS rebinding to bypass same-origin policy restrictions and send requests to the local MCP server. This could allow an attacker to invoke tools or access resources exposed by the MCP server on behalf of the user in those limited circumstances. Note that running HTTP-based MCP servers locally without authentication is not recommended per MCP security best practices. This issue does not affect servers using stdio transport. This vulnerability is fixed in 1.23.0.Show less

fastify-reply-from is a Fastify plugin to forward the current HTTP request to another server. Prior to 12.5.0, by crafting a malicious URL, an attacker could access routes that are not allowed, even though the reply.from is defined for specific routes in @fastify/reply-from. This vulnerability is fixed in 12.5.0.Show less

MCP TypeScript SDK is the official TypeScript SDK for Model Context Protocol servers and clients. Prior to 1.24.0, The Model Context Protocol (MCP) TypeScript SDK does not enable DNS rebinding protection by default for HTTP-based servers. When an HTTP-based MCP server is run on localhost without authentication with StreamableHTTPServerTransport or SSEServerTransport and has not enabled enableDnsRebindingProtection, a malicious website could exploit DNS rebinding to bypass same-origin policy restrictions and send requests to the local MCP server. This could allow an attacker to invoke tools or access resources exposed by the MCP server on behalf of the user in those limited circumstances. Note that running HTTP-based MCP servers locally without authentication is not recommended per MCP security best practices. This issue does not affect servers using stdio transport. This vulnerability is fixed in 1.24.0.Show less