Total
419404 CVE
| CVE | Vendors | Products | Updated | CVSS v2 | CVSS v3 |
|---|---|---|---|---|---|
| CVE-2026-58474 | 2026-09-24 | N/A | 8.8 HIGH | ||
| whichllm before 0.5.16 contains a code injection vulnerability in the run and snippet commands that allows a remote attacker who controls a HuggingFace repository to achieve arbitrary code execution by crafting a malicious GGUF filename containing double quotes or other special characters. The script generation function in cli.py interpolates HuggingFace-derived values, including GGUF variant filenames from the Hub API siblings rfilename field, directly into Python source code without escaping, allowing the crafted filename to break out of the generated string literal and execute injected code on the user's machine before any model download occurs. | |||||
| CVE-2026-85668 | 2026-09-24 | N/A | 7.5 HIGH | ||
| Xinference (affected commit 4a94832, v3.x) contains an unauthenticated arbitrary-path file read vulnerability in the POST /v1/models/llm/auto-register endpoint, which accepts a caller-supplied model_path parameter without authentication or path confinement. The endpoint reads and parses config.json, tokenizer_config.json, and chat_template.jinja files at the supplied path and reflects the parsed content back to the caller, allowing an unauthenticated attacker to probe the server filesystem and extract content of files with those names in any directory. | |||||
| CVE-2026-86121 | 2026-09-24 | N/A | 9.8 CRITICAL | ||
| Cua computer-server versions before 0.3.42 skip authentication when the CONTAINER_NAME environment variable is unset and bind to all interfaces by default, allowing unauthenticated attackers to execute arbitrary commands. Attackers can reach TCP port 8000 to run shell commands via the run_command endpoint, read and write arbitrary files through file operation endpoints, and access interactive PTY shells without authentication. | |||||
| CVE-2026-79784 | 2026-09-24 | N/A | 8.8 HIGH | ||
| Vocos instantiates a class named by a configuration file without restricting which class may be named. instantiate_class in vocos/pretrained.py takes the class_path value from the configuration, splits it into a module and an attribute, imports the module with __import__, resolves the attribute with getattr, and calls the result as args_class(*args, **kwargs) where kwargs is the config's own init_args mapping. No allowlist constrains the dotted path, so a configuration may name any importable callable and supply the arguments it is called with. Vocos.from_hparams reaches this for each of the feature_extractor, backbone and head entries, and Vocos.from_pretrained reaches it with a remote file: it downloads config.yaml from a caller-named Hugging Face repository and passes it straight to from_hparams. Loading a model from a repository the user does not control therefore executes code of the repository owner's choosing in the loading process. The neighbouring torch.load of the downloaded weights is a separate matter and is constrained on PyTorch releases that default weights_only to true, which leaves this path as the reachable one. | |||||
| CVE-2026-86123 | 2026-09-24 | N/A | 8.7 HIGH | ||
| SQL Chat contains four unauthenticated API endpoints that accept client-supplied database connection parameters and execute arbitrary SQL queries against attacker-specified hosts. Attackers can connect to internal databases, execute SQL commands, enumerate schemas, and pivot into the server's network without authentication. | |||||
| CVE-2026-90772 | 2026-09-24 | N/A | 7.6 HIGH | ||
| Amundsen frontend through 4.3.0 renders table, dashboard, and feature descriptions with dangerouslySetInnerHTML without HTML sanitization in ResourceListItem components. Attackers can inject malicious markup like img elements with onerror handlers into descriptions via the metadata service or Elasticsearch, executing JavaScript in every user's browser that views search results. | |||||
| CVE-2026-40526 | 2026-09-24 | N/A | 6.5 MEDIUM | ||
| Volmarg Personal Management System contains a path traversal vulnerability that allows authenticated attackers to read arbitrary files by supplying absolute filesystem paths to the GET /public/get-file/{path} endpoint. The path route parameter is passed directly to file_get_contents() without canonicalization against a permitted base directory, enabling attackers to retrieve sensitive files accessible to the PHP-FPM worker process without using directory traversal sequences. | |||||
| CVE-2026-84700 | 2026-09-24 | N/A | 8.6 HIGH | ||
| PikiwiDB (Pika) v3.5.7 exposes an internal protobuf replication server on a port derived from the client port plus 2000 (e.g. 11221 when the default client port 9221 is used) that does not authenticate incoming requests. Although requirepass is intended to gate replication — a slave presents it as masterauth inside its MetaSync request — only the MetaSync handler (HandleMetaSyncRequest) validates it; the frame dispatcher (DealMessage) does not require a completed or attempted MetaSync before routing other message types to their handlers. As a result, an unauthenticated remote attacker can connect directly to the replication port and issue TrySync, DBSync, BinlogSync, and RemoveSlaveNode requests, obtaining the full-sync snapshot and live write stream and removing replica nodes, even when requirepass is configured. | |||||
| CVE-2026-80049 | 2026-09-24 | N/A | 8.8 HIGH | ||
| Airbyte Platform resolves the workspace used for its authorization decision from a field the caller supplies. AuthorizationServerHandler copies recognised identifiers out of the raw JSON request body into X-Airbyte-* headers, and AuthenticationHeaderResolver.resolveWorkspace consults X-Airbyte-Workspace-Id ahead of every resource-derived header, including those for connection, source and destination identifiers. Endpoints whose declared request bodies carry only a resource identifier are nonetheless reached with an added workspaceId field, because the extractor reads the body rather than the endpoint's schema, so the permission check is performed against the workspace the caller nominated while the handler acts on the resource identifier the caller supplied. Nothing afterwards compares the resource's owning workspace with the one that was authorized. A member of any workspace can therefore read source and destination configuration, trigger and cancel syncs, and delete connections, sources and destinations that belong to workspaces they have no access to, at whatever privilege level their own workspace membership grants them. | |||||
| CVE-2026-5917 | 2026-09-24 | N/A | 8.8 HIGH | ||
| libgit2 versions before 1.8.7 and 1.9.0 before 1.9.7 built with the libssh2 SSH backend (USE_SSH=libssh2) contain a shell command injection vulnerability that allows remote attackers to execute arbitrary commands on an SSH server by supplying a repository path containing unescaped shell metacharacters such as single quotes, semicolons, or pipes. The gen_proto() function in ssh_libssh2.c inserts the repository path directly into a shell command string without escaping special characters before passing it to libssh2_channel_exec(), enabling an attacker to craft a malicious submodule URL in a .gitmodules file that, when processed during a recursive clone, causes the remote server's shell to interpret injected commands under the victim's SSH user account. | |||||
| CVE-2026-89425 | 2026-09-24 | N/A | 7.5 HIGH | ||
| UTF8DataInputJsonParser._reportInvalidToken() in FasterXML jackson-core builds the offending-token text for its error message by appending Java identifier characters to a StringBuilder in a loop that has no upper bound. Unlike the three sibling parser implementations, including UTF8StreamJsonParser, it never consults ErrorReportConfiguration.getMaxErrorTokenLength() (default 256). A malformed token supplied to a parser created through JsonFactory.createParser(DataInput) is therefore accumulated in full. No StreamReadConstraints setting mitigates this: maxDocumentLength cannot be applied to DataInput sources at all, and maxStringLength does not cover this path because the accumulation bypasses ReadConstrainedTextBuffer. The reporter measured a 20,000,109-character exception message from a 20-million-character malformed token on the DataInput path, against 367 characters for identical input on the InputStream path. Scaling the payload drives the StringBuilder, which also incurs byte-to-char expansion and internal array doubling, to many times the raw payload size and can trigger OutOfMemoryError for the whole JVM. UTF8DataInputJsonParser was introduced in 2.8.0 together with createParser(DataInput); releases before 2.8.0 do not contain the affected class. | |||||
| CVE-2026-85394 | 2026-09-24 | N/A | 9.1 CRITICAL | ||
| python-jose through 3.5.0 fails to properly validate asymmetric keys in HMAC initialization, accepting DER-encoded public keys that lack PEM armor or SSH prefixes. Attackers holding the service's public key can forge HS256 tokens that pass verification when algorithms are not explicitly restricted. This is an incomplete fix for CVE-2024-33663. | |||||
| CVE-2026-91922 | 2026-09-24 | N/A | 6.1 MEDIUM | ||
| Steedos Platform through 3.0.15-beta.47 contains a reflected cross-site scripting vulnerability in the anonymous /api/page/render endpoint that fails to properly escape query parameters in inline script elements. Attackers can craft malicious links with script-terminating sequences in the schemaApi or data parameters to execute arbitrary JavaScript in victim sessions and steal X-Auth-Token credentials. | |||||
| CVE-2026-81036 | 2026-09-24 | N/A | 8.1 HIGH | ||
| Stalwart Mail Server does not compare an OAuth redirect target against any registered destination in its default configuration. The validation routine in crates/http/src/auth/oauth/registration.rs returns success immediately when the client-authentication requirement is disabled, and that requirement is false in the shipped settings, so the supplied redirect value is neither matched against a registered client nor otherwise constrained. The value is stored with the authorization code, and the login page reads it back and sends the browser to it with the code attached. A request naming a destination the attacker controls therefore delivers a valid authorization code there once the account holder authenticates, and because the token endpoint checks only that the redirect presented at exchange matches the one recorded with the code, the same party can exchange it for access and refresh tokens and read the account's mail. | |||||
| CVE-2026-80050 | 2026-09-24 | N/A | 6.5 MEDIUM | ||
| ContiNew Admin fails to apply file-upload permission checks or file-type allowlist validation to multipart upload endpoints, allowing authenticated users to store files with arbitrary extensions. Attackers can initialize chunked uploads, send file parts, and complete uploads to leave arbitrary files in the storage backend accessible via web server URLs. | |||||
| CVE-2026-90944 | 2026-09-24 | N/A | 8.2 HIGH | ||
| Krayin CRM through 2.2.6 exposes the POST /admin/mail/inbound-parse endpoint without authentication, allowing unauthenticated attackers to inject arbitrary emails into the CRM inbox. Attackers can supply crafted RFC 2822 messages with forged sender information and headers to insert emails with any subject and body, including replies to existing conversation threads. | |||||
| CVE-2026-4637 | 2026-09-24 | N/A | N/A | ||
| Paessler PRTG Network Monitor before version 26.2.120.1449 is affected by a reflected Cross-Site Scripting (XSS) vulnerability. When a request is made for a non-existent resource ending in \".htm\", the web interface returns an HTTP 403 \"Forbidden Path\" error page that echoes the requested URL path into the HTML response body without proper output encoding or sanitization. An unauthenticated, remote attacker can craft a URL containing an HTML/JavaScript payload in the path (e.g. https:////welcome.htm) and, once a victim with an active PRTG session opens the crafted link, execute arbitrary JavaScript in the security context of the PRTG web interface. Because the PRTG session cookie is not protected with the HttpOnly attribute, successful exploitation allows the attacker to read and exfiltrate the victim's session cookie, potentially leading to session hijacking. | |||||
| CVE-2026-80427 | 2026-09-24 | N/A | 8.4 HIGH | ||
| bestzip builds the argument list for the system zip utility without separating options from operands. The destination archive path and the caller-supplied source paths are passed to the child process with no -- delimiter between them, so any source entry beginning with a hyphen is interpreted by zip as an option rather than a file name. zip accepts -T to test the finished archive and -TT to name the command used to perform that test, so a source list containing those two entries and a command string causes zip to run that command through a shell once the archive has been written. An application that passes a file name or path it received from an untrusted source into the bestzip API therefore executes a command of the supplier's choosing. Versions 2.2.6 and 3.0.2 add the delimiter. | |||||
| CVE-2026-80347 | 2026-09-24 | N/A | 7.5 HIGH | ||
| mcp-fetch checks a fetch target against its SSRF guard without removing the brackets that surround an IPv6 literal. isSafeUrl reads the hostname from the parsed URL, which for a literal such as http://[::1]/ yields the bracketed string, and then tests it with net.isIP. That call returns zero for a bracketed value, so the branch holding the private-address checks is skipped entirely. The guard falls back to resolving the hostname, the bracketed string is not a resolvable name, no addresses are returned, and the target is reported safe. The HTTP client then strips the brackets and connects. Because the address may be given in IPv4-mapped form, the same path reaches any IPv4 target the loopback and private checks were meant to exclude, including link-local metadata endpoints. isPrivateIPv6 also has no case for the ::ffff: prefix, so the mapped form would still pass even if the brackets were removed. The fetch target is supplied as a tool argument, so an attacker who can influence what the model requests can read internal responses back into the model context. | |||||
| CVE-2026-86116 | 2026-09-24 | N/A | 6.5 MEDIUM | ||
| Metabase versions before 0.63.1 fail to enforce data analyst permission checks on glossary API endpoints, allowing any authenticated user to create, modify, and delete glossary entries. Attackers can submit requests to POST, PUT, and DELETE glossary endpoints to tamper with instance-wide business glossary data without proper authorization. | |||||