SurrealDB versions from 2.2.0 up to but not including 4.0.0 expose record IDs of referencing rows in error messages when a delete operation is blocked by an ON DELETE REJECT reference. This exposure occurs without verifying the caller's SELECT permission on those referencing rows. As a result, an authenticated user with DELETE permission can infer the existence and exact IDs of referencing records that would otherwise be hidden by row-level SELECT restrictions.
SurrealDB versions 3.0.4 up to but not including 4.0.0 have a vulnerability in the GraphQL WebSocket upgrade handler at GET /graphql. The operator-configured limits for WebSocket frame, message, and buffer sizes are not enforced, causing the system to fall back to permissive library defaults. This allows unauthenticated remote attackers to send large amounts of data (up to 64 MiB per connection) by specifying the target namespace and database in request headers, potentially exhausting server memory and degrading availability.
SurrealDB versions 3.0.0 up to but not including 4.0.0 contain a path traversal vulnerability in the filesystem storage bucket. The vulnerability arises because the bucket root is canonicalized, but descendant symbolic links are not, allowing attackers to bypass lexical prefix checks. This enables unauthorized reading and writing of files across the host filesystem.
SurrealDB versions from 1.0.0 up to but not including 4.0.0 have a vulnerability in their embedded JavaScript engine that allows authenticated users to bypass capability checks. This occurs because the surrealdb.functions bridge calls internal dispatchers without enforcing function capability restrictions, enabling unauthorized execution of built-in SurrealQL functions.
SurrealDB versions from 3.1.0 up to but not including 4.0.0 contain an SQL injection vulnerability in the Model Context Protocol (MCP) server. The flaw arises because the identifier validator mishandles escaped backticks, allowing attackers to break out of quoted identifiers and execute arbitrary SurrealQL statements. This vulnerability has a high severity rating with a CVSS score of 8.2.
SurrealDB versions from 1.0.0 up to but not including 4.0.0 expose sensitive environment variable values (SURREAL_PASS and SURREAL_TOKEN) in clear text in the CLI help output. This occurs because the CLI commands do not hide environment variable values when displaying help, potentially exposing root database credentials to anyone who can view terminal outputs such as shell logs, CI logs, or screen shares.
SurrealDB versions 3.0.0 up to but not including 4.0.0 contain an authorization bypass vulnerability. This flaw allows users assigned the read-only VIEWER role to modify the database schema by implicitly creating new tables without proper permission checks. Specifically, issuing CREATE statements on nonexistent tables in non-STRICT mode adds arbitrary empty schemaless tables, potentially cluttering the schema or causing conflicts with application tables.
SurrealDB versions from 2.2.0 up to but not including 4.0.0 contain an authorization bypass vulnerability in the Document::purge_references function. This flaw allows authenticated users to bypass table-level delete and update permissions by exploiting REFERENCE ON DELETE cascades. As a result, attackers can delete or modify referencing records or execute custom expressions despite restrictive permissions.
SurrealDB versions 3.0.0 up to but not including 4.0.0 contain a permission bypass vulnerability. Authenticated users with low privileges can exploit the GET /export endpoint to read restricted records. This bypass allows attackers to access raw records without enforcement of table-level SELECT predicates and field-level SELECT permissions, exposing data that should be hidden.
SurrealDB versions prior to 4.0.0 have an authorization vulnerability in the filter_fields_by_permission function. This flaw allows authenticated users to read child field values that should be denied by SELECT permissions due to incorrect handling of permission checks on nested array elements.
SurrealDB versions prior to 4.0.0 have a guard bypass vulnerability that permits authenticated users to execute write subqueries within LIVE query WHERE clauses or field projections. This flaw enables injection of CREATE, UPDATE, DELETE, or RELATE statements that execute atomically inside other users' transactions. The vulnerability can lead to amplified write operations and unauthorized exfiltration of document data even after permissions have been revoked.
SurrealKV versions before 1.1.0 have an out-of-bounds read vulnerability in the InternalKeyRef::from_encoded constructor. This constructor only performs key length checks using debug-only assertions, which are not enforced in production. If callers provide keys shorter than 16 bytes or if SSTable blocks are crafted or corrupted, this can cause unchecked length underflow in functions like timestamp(), seq_num(), and trailer(), leading to reading memory before the allocated buffer or causing a panic.
SurrealDB versions 3.0.0 up to but not including 4.0.0 contain an authorization bypass vulnerability. This flaw allows clients with table write access to persist writes that should be refused by exploiting THROWing table events within interactive transactions. Attackers can execute guarded statements via RPC begin and query, ignore errors, and commit changes to persist unauthorized records and side effects.
Vikunja versions from 0.8.0 up to but not including 2.8.0 contain a denial of service vulnerability. Authenticated project readers can exploit this by sending CalDAV PROPFIND requests with repeated properties, causing excessive resource consumption on the server. This leads to exhaustion of memory, CPU, and bandwidth, resulting in denial of service.
Vikunja versions from 2.2.0 up to but not including 2.8.0 have an information disclosure vulnerability. Authenticated users can exploit the tasks.overdue webhook to read private project details of other users, including project titles, descriptions, and owner IDs. This occurs by registering a personal tasks.overdue webhook with a subscribed overdue task and receiving sensitive data at external URLs.
OpenRemote core versions before 1.34.0 have an authentication bypass vulnerability in the UserAssetProvisioningMQTTHandler component. This flaw allows unauthenticated attackers to impersonate provisioned devices by publishing a known device's public certificate to the provisioning request topic. As a result, attackers can authenticate their anonymous MQTT connection as the device's service account, enabling them to read and modify assets.
Vikunja versions from 0.20.5 up to but not including 2.8.0 contain a login cross-site request forgery (CSRF) vulnerability in the OpenID Connect callback handler. This flaw allows attackers to craft a callback link with an empty state parameter and their own authorization code, causing victims to be silently logged into the attacker's account and potentially exposing their entered data.
Vikunja versions from 2.3.0 up to but not including 2.8.0 contain a vulnerability in the CSV migration module where uncontrolled memory allocation can occur. This happens because the module does not de-duplicate label column mappings or limit the number of labels per row before building the full import. Authenticated users can exploit this by sending a specially crafted request to the migration endpoint, causing the server to exhaust memory and potentially crash due to out-of-memory conditions.
Vikunja versions from 2.6.0 up to but not including 2.8.0 contain a resource exhaustion vulnerability in the Planka migration converter. This flaw allows an attacker controlling a Planka server to supply boards with duplicate list IDs and many cards, causing Vikunja to multiply task allocations excessively. The resulting memory exhaustion can lead to the Vikunja process being terminated by the operating system (OOM-killed).
Vikunja versions from 0.16.0 up to but not including 2.8.0 have an information disclosure vulnerability. This flaw causes users' Trello API tokens to be written verbatim to application logs and Sentry error telemetry when Trello migration fails due to transport errors. The token is embedded in the URL error, exposing reusable third-party credentials to anyone with access to logs or telemetry data.