CVE Watch
Every published CVE, mapped to engagement reality.
Crawled from cve.org every day. Each entry annotated with the QSearch coverage signal — how many of our agents, skills, and playbooks address the technique. Subscribe via RSS for SIEM pipe, or get the weekly digest by email.
A vulnerability in the LightGlue model loading path of huggingface/transformers version 5.2.0 allows an attacker-controlled model reposit...
A vulnerability in the LightGlue model loading path of huggingface/transformers version 5.2.0 allows an attacker-controlled model repository to execute arbitrary code during model initialization. The issue arises because the `trust_remote_code` parameter, intended to prevent remote code execution, is overridden by untrusted serialized configuration data in a nested code path. Specifically, when loading a LightGlue model using `AutoModel.from_pretrained()` with `trust_remote_code=False`, the `LightGlueConfig` reads the `trust_remote_code` value from the untrusted `config.json` file and propagates it into nested `AutoConfig.from_pretrained()` calls. This results in the execution of attacker-provided Python modules, even when the victim explicitly disables remote code execution. The vulnerability poses a high risk for environments such as API inference servers, research notebooks, CI/CD pipelines, and model evaluation workers, potentially leading to credential theft, lateral movement, or persistence/backdoor deployment.
huggingfaceCWE-829An unauthenticated remote attacker can recover a default, hard coded password from a firmware image and thus gain full access to all affe...
An unauthenticated remote attacker can recover a default, hard coded password from a firmware image and thus gain full access to all affected devices.
mbs-solutionsCWE-1393ZDRES-232: resolveProxyClass Not Overridden - acceptMatchers Filter Bypass via java.lang.reflect.Proxy Assessment: Fully addressed
ZDRES-232: resolveProxyClass Not Overridden - acceptMatchers Filter Bypass via java.lang.reflect.Proxy Assessment: Fully addressed. When the serialised stream contains a TC_PROXYCLASSDESC (the marker for a java.lang.reflect.Proxy ), JDK’s ObjectInputStream.readProxyDesc() is dispatched. JDK then calls the default ObjectInputStream.resolveProxyClass(interfaces) implementation, which performs Class.forName(intf, false, latestUserDefinedLoader()) for EACH interface name and constructs the proxy class — bypassing the accepted classes list . ZDRES-233: Class.forName(name, initialize=true, classLoader) in readClassDescriptor Triggers Static Initialiser of Allow-Listed Classes Assessment: Fully addressed. For ANY class on the allow-list, deserialising a stream that names it triggers the class’s (static initialiser) BEFORE any instance is constructed. This means an attacker who supplies a class name on the allow-list (e.g., the developer wrote accept(“com.myapp.*") , attacker supplies com.myapp.SomeClass ) causes <clinit> of SomeClass — and many real-world classes have side-effecting static initialisers Both issues have been fixed.
CWE-502Files or directories accessible to external parties vulnerability in ABB T-MAC Plus
Files or directories accessible to external parties vulnerability in ABB T-MAC Plus. This issue affects T-MAC Plus: 4.0-24.
abbCWE-552LibreChat is an enhanced ChatGPT clone that supports multiple AI providers
LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. In versions up to and including 0.8.3, the Model Context Protocol (MCP) server integration resolves ${VAR} placeholders against the server's process.env during Zod schema validation of user-supplied MCP server URLs. Any authenticated user can create a malicious MCP server configuration with a URL pointing to an attacker-controlled domain containing environment variable references, causing the LibreChat server to connect to the attacker's server and transmit critical secrets such as CREDS_KEY, CREDS_IV, JWT_SECRET, and MONGO_URI in the request URL. This enables full compromise of the installation's cryptographic materials and database credentials without requiring administrative privileges. This is patched in version 0.8.4-rc1.
librechatCWE-200authentik is an open-source identity provider
authentik is an open-source identity provider. Prior to versions 2025.12.6, 2026.2.4, and 2026.5.1, the Source stage can be bypassed by sending an empty POST. This issue has been patched in versions 2025.12.6, 2026.2.4, and 2026.5.1.
goauthentikCWE-287authentik is an open-source identity provider
authentik is an open-source identity provider. Prior to versions 2025.12.5 and 2026.2.3, due to the implementation of stages in the SFE (Simple Flow Executor) in order to make the interface more compatible with legacy browsers, it was possible to use an XSS exploit in the AutosubmitStage. This issue has been patched in versions 2025.12.5 and 2026.2.3.
goauthentikCWE-79The ARMember Premium plugin for WordPress is vulnerable to an insecure password reset mechanism in all versions up to, and including, 7.3.1
The ARMember Premium plugin for WordPress is vulnerable to an insecure password reset mechanism in all versions up to, and including, 7.3.1. The plugin stores a plaintext copy of the password reset key in the `arm_reset_password_key` user meta field when a user requests a password reset. This is in addition to the hashed key that WordPress core stores securely in `wp_users.user_activation_key`. The plaintext key stored in `wp_usermeta` can be used with the plugin's custom `armrp` reset action to set a new password for any user. Combined with another vulnerability such as SQL Injection (CVE-2026-5073, CVE-2026-5074), this makes it possible for unauthenticated attackers to extract the plaintext reset key and take over any user account, including administrators.
CWE-287CrowCpp Crow through v1.3.1 HTTP is vulnerable to response header injection via unvalidated response header values
CrowCpp Crow through v1.3.1 HTTP is vulnerable to response header injection via unvalidated response header values.
CWE-113OpenClaude is an open-source coding-agent command line interface for cloud and local model providers
OpenClaude is an open-source coding-agent command line interface for cloud and local model providers. Prior to version 0.5.1, the dangerouslyDisableSandbox parameter is exposed as part of the BashTool input schema, meaning the LLM (an untrusted principal per the project's own threat model) can set it to true in any tool_use response. Combined with the default allowUnsandboxedCommands: true setting, a prompt-injected model can escape the sandbox for any arbitrary command, achieving full host-level code execution. This issue has been patched in version 0.5.1.
gitlawbCWE-284CWE-306Spacelabs Healthcare Sentinel versions 10.5.x and higher and 11.x.x before 11.6.0 contain an unauthenticated remote code execution vulner...
Spacelabs Healthcare Sentinel versions 10.5.x and higher and 11.x.x before 11.6.0 contain an unauthenticated remote code execution vulnerability through a deprecated .NET Remoting HTTP channel exposed on port 8989 that allows attackers to perform arbitrary file read and write operations by supplying valid .NET URI endpoints. Attackers can write ASPX webshells to the IIS wwwroot directory to achieve unauthenticated remote code execution on the system. Port 8989 is not exposed in a default Sentinel installation; exploitation requires that the .NET Remoting port has been explicitly made network-accessible through deliberate configuration or network policy changes.
CWE-306OpenMed before 1.5.2 contains a remote code execution vulnerability in the PII privacy-filter model loading path
OpenMed before 1.5.2 contains a remote code execution vulnerability in the PII privacy-filter model loading path. The privacy-filter dispatcher used broad substring matching on the user-supplied model_name parameter, allowing a value such as attacker/foo-privacy-filter-bar to route through a path that loads Hugging Face models with trust_remote_code=True. An unauthenticated attacker can supply a malicious model repository containing custom Transformers code via auto_map in config.json or tokenizer_config.json, which is imported and executed with the privileges of the OpenMed service process.
CWE-94CWE‑522: Insufficiently Protected Credentials in web services in Progress Sitefinity version from 14.0.7700 to 14.4.8152, and 15.0.8200 t...
CWE‑522: Insufficiently Protected Credentials in web services in Progress Sitefinity version from 14.0.7700 to 14.4.8152, and 15.0.8200 to 15.0.8234, and 15.1.8300 to 15.1.8335, 15.2.8400 to 15.2.8441, 15.3.8500 to 15.3.8531, and 15.4.8600 to 15.4.8630 allows a remote unauthenticated attacker to obtain plain-text credentials used connect to Sitefinity Insight service. Successful exploitation requires active integration with Sitefinity Insight and non-default site configuration.
progressCWE-522CWE-284: Improper Access Control in web services in Progress Sitefinity 15.4.8623 before 15.4.8630 allows a remote unauthenticated attack...
CWE-284: Improper Access Control in web services in Progress Sitefinity 15.4.8623 before 15.4.8630 allows a remote unauthenticated attacker to access content that should be restricted, resulting in full compromise of confidentiality, integrity, and availability of affected installations.
progressCWE-284Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in Ahmad WP Job Portal allows Blind SQ...
Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in Ahmad WP Job Portal allows Blind SQL Injection. This issue affects WP Job Portal: from n/a through 2.5.1.
CWE-89Incorrect Privilege Assignment vulnerability in Themeisle Masteriyo LMS PRO allows Privilege Escalation
Incorrect Privilege Assignment vulnerability in Themeisle Masteriyo LMS PRO allows Privilege Escalation. This issue affects Masteriyo LMS PRO: from n/a through 2.20.0.
CWE-266The Kirki – Freeform Page Builder, Website Builder & Customizer plugin for WordPress is vulnerable to privilege escalation via account ta...
The Kirki – Freeform Page Builder, Website Builder & Customizer plugin for WordPress is vulnerable to privilege escalation via account takeover in all versions 6.0.0 to 6.0.6. This is due to the plugin accepting an arbitrary email address when a username is used in the password reset request. This makes it possible for unauthenticated attackers to send a password reset link for any user registered on the site to their own email address.
CWE-269Langroid is a framework for building large-language-model-powered applications
Langroid is a framework for building large-language-model-powered applications. Prior to version 0.63.0, SQLChatAgent executes SQL produced by an LLM, which is influenceable by prompt injection. When configured with a database role that has privileges enabling code execution or filesystem access (e.g., PostgreSQL pg_execute_server_program, MySQL FILE, MSSQL xp_cmdshell), an attacker who can shape the agent's input — including indirectly via data returned to the LLM — can coerce execution of dialect-specific primitives such as `COPY ... FROM PROGRAM`, achieving RCE on the database host. Fixed in v0.63.0 by defaulting SQLChatAgent to a SELECT-only sqlglot-parsed statement allowlist with a dialect-aware dangerous-pattern blocklist; allow_dangerous_operations=True restores the previous unrestricted behavior for trusted deployments.
CWE-89CWE-94Cloud Foundry UAA versions v76.12.0 through v78.12.0 are vulnerable to a private key exposure
Cloud Foundry UAA versions v76.12.0 through v78.12.0 are vulnerable to a private key exposure. The server contains a vulnerability where EC (Elliptic Curve) private keys are inadvertently exposed through the public /token_keys endpoint. This endpoint is designed to provide public key material for JWT token verification but incorrectly exposes private key components for EC keys. The vulnerability affects deployments using EC keys for JWT token signing. The vulnerability does not affect RSA key configurations, only deployments using EC keys for JWT signing. Affected versions: - uaa_release: v76.12.0 through v78.12.0 (inclusive); fixed in v78.13.0 or later - CF Deployment: v30.0.0 through v56.0.0 (inclusive); fixed in v56.1.0 or later (bundles uaa_release v78.13.0)
CWE-200Arm Whois 3.11 contains a stack-based buffer overflow vulnerability that allows remote attackers to execute arbitrary code by supplying o...
Arm Whois 3.11 contains a stack-based buffer overflow vulnerability that allows remote attackers to execute arbitrary code by supplying oversized input to the IP address or domain field. Attackers can craft malicious input exceeding 658 bytes with shellcode to overwrite the structured exception handler and gain command execution when the application processes the input.
CWE-121
9.6 CRITICALWeekly digest
Get the curated CVE digest every Monday
One email a week, sent Monday morning CET. The CVEs published or modified in the last seven days, severity-ordered, with the QSearch coverage signal. Unsubscribe with one click — included in every send.
Pipe the CVE feed into your stack.
CVE Watch publishes RSS, Atom, and JSON feeds — wire them into your SIEM, Slack, Discord, or your RSS reader of choice. Or get the curated weekly digest by email.