Intelligence
Explorer.

docling-core 2.5.0 before 2.103.0 contains a local file disclosure vulnerability that allows attackers to read local images by supplying plain filesystem paths in ImageRef URIs, bypassing the allow_image_file_uri gate. Attackers can submit crafted DoclingDocument JSON or YAML that, when exported with ImageRefMode.EMBEDDED or rendered to HTML, embeds readable local image files as base64 and reveals path existence.

@rvf/set-get before 7.0.4 contains a prototype pollution vulnerability in setPath that allows unauthenticated attackers to overwrite properties of shared built-in functions via inherited Object.prototype methods. Attackers can submit a form field named hasOwnProperty.call through @rvf/core form parsing to replace the shared function property, causing process-wide denial of service until restart.

crewai-tools through 1.15.28 contains a server-side request forgery vulnerability in SeleniumScrapingTool that allows attackers to fetch internal resources because it never applies the validate_url private-address check. Attackers influencing agent tool input through prompt injection can direct headless Chrome to loopback, RFC1918, or cloud metadata addresses to disclose internal data and cloud credentials.

sqlparse through 0.6.0 contains a PHP code injection vulnerability that allows attackers to inject PHP expressions by supplying SQL text containing unescaped dollar signs. Attackers can embed ${expr} or {${expr}} sequences that OutputPHPFilter places in double-quoted strings, executing arbitrary PHP code when the generated snippet runs.

NocoBase @nocobase/plugin-collection-sql through 2.2.28 contains an SQL injection vulnerability that allows privileged users to store unvalidated SQL by bypassing checkSQL() through the sqlCollection:setFields action. Attackers holding the collection-sql snippet can replace a benign collection query with dangerous statements executed on reload, reading pg_shadow password hashes or calling pg_read_file.

Semaphore UI through 2.19.16 contains a remote code execution vulnerability that allows users with the task_runner role to override protected Ansible variables by supplying arbitrary keys in the task environment JSON. Attackers can POST tasks setting ansible_ssh_common_args with a ProxyCommand or ansible_connection=local to execute commands as the Semaphore server user, exposing encryption keys, SSH keys and stored secrets.

docTR through 1.1.0 contains an unsafe deserialization vulnerability in load_pretrained_params() that loads checkpoints with torch.load without weights_only=True. Attackers who control a local or URL-supplied checkpoint can embed a malicious pickle payload to execute Python code with the loading process's privileges on torch releases before 2.6.

GDevelop desktop app through 5.6.284 contains a credential leak vulnerability that allows attackers to steal the gd_resource cookie by supplying attacker-controlled resource URLs in project files. Attackers can get victims to open a crafted project, causing downloadLocalFile() to send the cookie to their server and expose private cloud project resources.

A weakness has been identified in MikroTik RouterOS up to 6.40.5/6.44/7.20.8. Impacted is the function same_check of the file ipt_SAME.ko. Executing a manipulation can lead to integer overflow. The attack may be launched remotely. Upgrading to version 7.21.5 and 7.23.1 is recommended to address this issue. You should upgrade the affected component.

A weakness has been identified in MemTensor memmy-agent up to 1.0.9/1.1.1/1.2.2. The impacted element is the function requireAdminWrite of the file Memory/src/server/http.ts of the component Memory Service. This manipulation causes resource consumption. Remote exploitation of the attack is possible. The attack is considered to have high complexity. The exploitability is regarded as difficult. The exploit has been made available to the public and could be used for attacks. The exploit is partially overstated. While "POST /api/v1/admin/shutdown" accepts unauthenticated requests when no token is configured, it’s designed for local use and later received cross-origin fixes. Unauthenticated endpoint exposure is confirmed; browser-triggered DoS remains likely but depends on specific browser behavior.

A vulnerability was detected in vraj404 Criminal-Management-System up to 2808d057e7ddc5295519a487365b0cacfd0ba8c4. This affects the function mysqli_error of the file Officers/addOfficer.php of the component Officer Registration Form. Performing a manipulation of the argument offName/offID/ID/contact/gender/weapon/role results in sql injection. It is possible to initiate the attack remotely. The exploit is now public and may be used. This product is using a rolling release to provide continious delivery. Therefore, no version details for affected nor updated releases are available. The project was informed of the problem early through an issue report but has not responded yet.

A security vulnerability has been detected in vraj404 Criminal-Management-System up to 2808d057e7ddc5295519a487365b0cacfd0ba8c4. Affected by this issue is the function mysqli_query of the file Officers/searchOff.php of the component Search Page. Such manipulation of the argument Search leads to sql injection. The attack may be performed from remote. The exploit has been disclosed publicly and may be used. This product utilizes a rolling release system for continuous delivery, and as such, version information for affected or updated releases is not disclosed. The project was informed of the problem early through an issue report but has not responded yet.

Mercury is a Python framework for sharing Jupyter notebooks as web applications and dashboards. Prior to 3.2.6, default tokenless deployments allow anonymous clients to create Jupyter kernels and send execute_request messages containing arbitrary Python. The endpoint blocklist does not cover /api/kernels, and the websocket wrapper previously tracked idle activity without authorizing executable messages. This issue is fixed in version 3.2.6.

Mercury is a Python framework for sharing Jupyter notebooks as web applications and dashboards. From 3.0.0 until 3.2.4, a missing comma in BLOCKED_PATTERNS concatenates the /lab and /api/terminals regular-expression literals. Requests to /api/terminals therefore bypass the anonymous endpoint blocklist and reach Jupyter Server's terminal API. An unauthenticated attacker can use the resulting interactive operating-system shell to execute commands and access data available to the Mercury process. This issue is fixed in version 3.2.4.

FOSSBilling is an open source billing and client-management platform. From 0.1.0 until 0.8.7, the unauthenticated POST /api/guest/client/create endpoint reveals whether a submitted email belongs to a registered client because guest signup returns a distinct unredacted error when the submitted email already exists. The trigger is an unauthenticated request to POST /api/guest/client/create with a target email address. An attacker can compare the guest-signup response for a target email to determine whether the email belongs to an existing FOSSBilling client. Box\Mod\Client\Api\Guest::create returns the error after Client\Service::clientAlreadyExists finds a matching record, and the API response exposes that exception message without redaction. This issue is fixed in version 0.8.7.

FOSSBilling is an open source billing and client-management platform. From 0.1.0 until 0.8.7, the unauthenticated api/guest/system/template_exists endpoint permits path traversal because templateExists uses Path::join on the attacker-controlled file parameter without rejecting .. sequences or verifying that the resolved path remains inside the active theme directory. The trigger is an unauthenticated request to guest/system/template_exists containing traversal sequences in the file parameter. An attacker can submit traversal sequences in the file parameter and observe the returned boolean to determine whether arbitrary files or directories exist on the server. The oracle does not disclose file contents. This issue is fixed in version 0.8.7.

FOSSBilling is an open source billing and client-management platform. From 0.8.4 until 0.8.6, hasManagePermission fails open for inactive non-core modules because its early return on !isExtensionActive skips the module-access and manage_settings checks. A low-privileged staff account with module-level extension access targets a deactivated non-core module through extension/config_get or extension/config_save. The account can call api/admin/extension/config_get or api/admin/extension/config_save for a deactivated non-core module to read decrypted retained extension secrets and overwrite stored extension configuration. The getConfig function returns the decrypted stored configuration despite the module's inactive state. This residual path remained after the earlier CVE-2026-62962 authorization fix. This issue is fixed in version 0.8.6.

CLI Bridge provides filesystem access within a workspace root. Prior to 0.0.2, a dangling symlink pointing outside the workspace root bypasses resolveSafePath when fs.realpathSync throws ENOENT for a nonexistent target. The fallback resolves only the parent directory and returns the symlink path as safe, so write_file follows the link and can write arbitrary files outside the workspace root. This issue is fixed in version 0.0.2.

Camunda 7.24.0 before 7.24.14 and 7.23.0 before 7.23.17 contains an authorization bypass vulnerability in the Cockpit webapp that allows authenticated attackers to bypass permission checks due to an inverted comparison logic in AbstractCockpitPluginResource.isPermissionDisabled(). Attackers with minimal Cockpit access can exploit a non-empty disabledPermissions configuration to read all process instances, process definitions, and incidents engine-wide without triggering any logging, regardless of their actual data grants.

Camunda 7.24.0 before 7.24.14 and 7.23.0 before 7.23.17 contains an expression injection vulnerability where AdhocQueryValidator.validate() fails to check sub-query expression maps within the orQueries field exposed by TaskQueryDto, allowing authenticated users to bypass the enableExpressionsInAdhocQueries restriction. Attackers can supply a malicious JUEL expression such as dueAfterExpression inside an orQueries entry to achieve arbitrary Java reflection and invoke java.lang.Runtime.exec, resulting in operating-system command execution as the engine service user.

Camunda 7.24.0 before 7.24.14 and 7.23.0 before 7.23.17 contains a missing authorization check vulnerability in engine-rest task sub-resources that allows authenticated attackers with no permission grants to access task comments and attachments by calling singular read and delete endpoints that lack the authorization-filtered checks enforced on their corresponding list endpoints. Attackers can exploit TaskAttachmentResourceImpl and TaskCommentResourceImpl methods such as getAttachment, getAttachmentData, deleteAttachment, and getComment using only a known object ID to read, download, or permanently delete task attachments and comments.

Incorrect default permissions in the Windows installer of Progressive Robot hMailServer 6.0.0 through 6.3.6 allow a local authenticated user to run code as the mail server's high-privileged accounts when the product is installed in a folder whose permissions are inherited from a writable parent, such as the root of a drive. The installer places the server's programs in the Bin folder of the program folder and gives neither the program folder nor Bin any permissions of their own, so they inherit the parent's. Under the default location, Program Files, ordinary users receive only read and execute and the configuration is not affected; but the installer accepts any path without warning, and a program folder at the root of an NTFS volume inherits write access for a broad group - Authenticated Users (Modify) on a Windows client, Users (create files and folders) on Windows Server. A local user without administrator rights can then alter or replace the service program, which runs as LocalSystem or as the service's virtual account, or the programs an administrator runs elevated from Bin during setup and maintenance, or place a library in Bin that those programs load in place of the system copy, gaining code execution with those privileges. The default Program Files installation is not affected.

Unchecked input for a loop condition in Progressive Robot hMailServer 6.0.0 through 6.3.6 allows a remote unauthenticated attacker to stop the server carrying mail. The server bounds the mail it generates in response to a message - a rule forward, reply or copy, an account forward, a Sieve redirect, vacation or notify, a bounce, an out-of-office reply - by a counter it carries in the message's X-hMailServer-LoopCount header, and makes no more once the counter reaches the configured rule loop limit (5 by default); that is what ends a forwarding or auto-reply cycle in a bounded number of hops. X-hMailServer-LoopCount is an ordinary header that is not stripped from mail arriving from outside, and the server read it with a signed conversion and no validation, so a count below zero passed every test against the limit: a sender who sets X-hMailServer-LoopCount to a negative value, or on the Linux builds to a value above the 32-bit signed range that the conversion wraps to a negative one, lifts the bound, and a message that enters a generated-mail cycle then loops up to about two thousand million times, each hop a queued message and a delivery, filling the queue and disk and stopping mail service. The Sieve notify action (from 6.2.22) consulted no limit at all, and its increment of the count could overflow a signed integer; and an account could reset the counter on its own mail through a Sieve editheader of the header (from 6.2.22) or a rule Set header value action (all versions) ahead of a forward or redirect.

Missing authorization in the Sieve notify action (6.2.22 through 6.3.6) and in the mail-rule Reply action (6.0.0 through 6.3.6) of Progressive Robot hMailServer allows an authenticated user to send mail through the server as any address. A Sieve notify sends a notification message, and its :from tag names the sender; the server set the notification's From header to that :from as the script wrote it, without checking that the address belongs to the account - the check that the Sieve vacation action applies to its own :from, and that the server applies to the sender of mail queued over COM or sent from the webmail. An account that can store a Sieve script (over ManageSieve, the REST API's self-service filter, or the administration interface) could therefore have the server send a notification from any address, on this server or anywhere; and where the claimed address is at a domain the server hosts with DKIM signing enabled, the server signed the notification with that domain's DKIM key, because it signs by the From header's domain, producing forged mail aligned to pass the domain's DMARC policy. The notify action and this behaviour were introduced in 6.2.22. In the mail-rule engine, a rule's Reply was likewise sent from whatever address the action named, whoever the rule belonged to, so that an account acting on its own rules over COM, a domain administrator, or an API key confined to a domain (over the REST API from 6.3.4) could have a domain's or an account's rule reply from any address, signed, where that address is at a domain the server hosts with DKIM signing enabled, with that domain's key.