CVE Vulnerability Catalog

LobsterAI 2026.5.27 through 2026.9.23 contains an external control of file path vulnerability in the skills:delete IPC handler that trusts the openclawSourceDir value from a skill's _meta.json during uninstall. Attackers who convince a user to install a crafted skill can make uninstallation recursively delete arbitrary user-writable directories, such as the home directory.

MariaDB server from 10.6.1 until 10.6.28, 10.11.19, 11.4.13, 11.8.9, 12.3.3, and 13.0.2 contains a vulnerability in the CONNECT engine's DOS table type where an incorrect boundary check permits a one-byte null write beyond a stack buffer at an attacker-controlled offset. An authenticated user able to use the CONNECT engine could cause a crash and potentially remote code execution.

A race condition vulnerability in the SessionManager of Argo CD v3.0.6 allows attackers to bypass rate limiting and perform a brute force attack via repeated crafted requests.

Hayat Health Facilities Inc. (Hayat Hospital) Hayat Mobile from 3.3.0 before 3.4.0 contains a use of client-side authentication vulnerability that allows authentication bypass.

EIPStackGroup OpENer v2.3.0/master up to commit 76b95cf contains an out-of-bounds read vulnerability in Connection Manager request parsing. This allows a remote attacker to cause a denial of service.

An out-of-bounds read vulnerability exists in EIPStackGroup OpENer v2.3 and master up to commit 76b95cf in the EtherNet/IP TCP SendRRData Common Packet Format parser. The issue occurs in CreateCommonPacketFormatStructure() when parsing optional socket address items of type 0x8000 or 0x8001 without validating that the remaining CPF buffer contains the complete fixed sockaddr structure, allowing a remote attacker to cause a denial of service.

An out-of-bounds read vulnerability exists in EIPStackGroup OpENer v2.3 and master through commit 76b95cf in the server-side CIP SetAttributeList service. This allows a remote attacker to cause a denial of service.

OpENer v2.3.0 / commit 76b95cf contains an out-of-bounds read in the unconnected explicit messaging path. This allows a remote attacker to cause a denial of service.

ILIAS before 9.24, 10.12, and 11.5 contains an unrestricted file upload vulnerability in QTI question import image handling (ilQtiMatImageSecurity) that allows authenticated authors to write executable files. Attackers with question pool import rights can import a crafted archive writing a .htaccess and PHP file to the web-served image directory, achieving remote code execution as the web server user.

MariaDB server from 10.6.1 until 10.6.28, 10.11.19, 11.4.13, 11.8.9, 12.3.3, and 13.0.2 RPM packages created the dedicated mysql service account with the database data directory as its home directory. A database user with the FILE privilege could write startup dot-files such as .bash_profile into $HOME, which could execute when an administrator opened a login shell for the mysql account. Debian packages are not affected because they use /nonexistent as the account home.

In Nginx UI from 2.0.0 until 2.5.0, the api/cluster router exposes node and namespace mutation operations and cluster-wide Nginx reload or restart operations with AuthRequired but without RequireSecureSession. An authenticated OTP-enabled user possessing a stolen or persisted JWT can perform node CRUD, read or replace node credentials, change namespaces, and invoke reload/restart without a fresh second-factor step-up.

In Nginx UI from 2.0.0 until 2.5.0, the self-upgrade mechanism validates a downloaded binary only with a same-origin digest obtained from the same upgrade mirror. A compromised mirror or network attacker able to alter both responses can supply a malicious executable and matching digest, and the application installs and runs the attacker-controlled code in the Nginx UI process context on the next upgrade.

In Nginx UI from 2.0.0 until 2.5.0, ordinary authenticated users can access /api/nodes and /api/nodes/:id, whose responses serialize the node token field. The same token is accepted as X-Node-Secret by AuthRequired and maps the request to initUser, allowing the user to impersonate a trusted node against a reachable cluster member.

In Nginx UI from 2.0.0 until 2.5.0, internal/backup/restore.go extracts inner archives before applying the restore_nginx and restore_nginx_ui flags and permits symlinks targeting the live Nginx configuration path. An authenticated user who can create and restore backups can craft a valid backup that places a symlink in the staging tree and then writes a regular file through that link, even when both restore flags are false.

In Nginx UI from 2.0.0 until 2.5.0, AuthRequired accepts a browser-managed token cookie as an API credential after the front end stores the JWT in that cookie. Because management endpoints do not universally require a CSRF token or perform Origin or Referer validation, a remote attacker can induce a logged-in administrator's browser to submit authenticated cross-site state-changing requests, including POST /api/configs.

In Nginx UI from 2.0.0 until 2.5.0, POST /api/login checks EnabledOTP but does not require a WebAuthn assertion when EnabledPasskey is true and no TOTP secret is configured. A passkey-only account is therefore issued a session after password verification, despite Enabled2FA reporting that the account has a second factor.

In Nginx UI from 2.0.0 until 2.5.0, Nginx UI accepts the Node.Secret master credential through the node_secret query parameter in HTTP and WebSocket authentication paths instead of requiring the X-Node-Secret header. The credential can consequently appear in access logs, proxy logs, browser history, Referer headers, configuration URLs, and deployment environment data.

SmarterMail before build 9777 contains a privilege escalation vulnerability where JWT access and refresh tokens embed a role claim at issuance that is not revalidated against the account's current role when redeemed through POST /api/v1/auth/refresh-token. Attackers who capture a refresh token issued before an administrator demotion, or a demoted user whose session was not actively polling at the time of demotion, can replay the stale token to obtain a new access token retaining the higher-privilege role.

SmarterMail before build 9777 contains a remote code execution vulnerability that allows an attacker holding a SysAdmin-scoped access token to bypass the Volume Mount script-directory containment control by provisioning a new mail domain with an arbitrary FileStore root path inside the trusted Scripts directory via the domain-put endpoint. Attackers can disclose the Scripts path through the AddOrUpdateMount endpoint, clear the upload extension blacklist via the global-mail endpoint, then upload a malicious script through the ordinary mail file-storage upload API so that saving a CommandMount triggers RunScript before validation, resulting in a reverse shell executing as the SmarterMail service account with SYSTEM-level privileges.

Allocation of resources without limits or throttling (CWE-770) during Java object deserialization in Google Guava versions 4.0 through 33.7.1 allows an attacker to cause a Denial of Service via OutOfMemoryError. When deserializing CompactHashMap, CompactHashSet, or MapMakerInternalMap instances, Guava eagerly allocates an array based on a caller-specified size parameter without throttling, permitting memory exhaustion from crafted serialization streams.