Base de CVEs

Improper Access Control (CWE-284) in Kibana can lead to unauthorized modification of Entity Analytics Watchlist configuration and potential information disclosure. A low-privileged authenticated user with read-only Security Solution access could perform write operations on watchlist data that should require elevated privileges. Under specific deployment conditions, this could also allow such a user to access data beyond their authorized scope.

Incorrect Authorization (CWE-863) in Elasticsearch can allow an authenticated user with limited index privileges to exploit insufficient authorization controls in the ingest simulation feature. By targeting indices they are not authorized to access directly, the user can cause those indices' configured ingest pipelines to execute and return their output, potentially disclosing data processed or enriched by those pipelines. Additionally, the same feature can be used to retrieve index mapping metadata for indices the user are not authorized to access directly.

Unintended Proxy or Intermediary ('Confused Deputy') (CWE-441) in Kibana can lead to unauthorized information exposure via Accessing Functionality Not Properly Constrained by ACLs (CAPEC-1). Under certain conditions, a lower-privileged user can cause data from sources they are not authorized to access to be processed using another user's privileges.

nono is software that allows users to run AI agents in a zero-latency sandbox. Prior to version 0.55.0, the nono Landlock/seccomp policies allow access to local Unix domain sockets (concrete and abstract). This allows an easy sandbox escape by talking to the per-user systemd dbus socket. Version 0.55.0 patches the issue.

The `@ai-sdk/harness-opencode` tool connects HarnessAgent to OpenCode through a sandboxed bridge. Prior to version 1.0.28, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (`host-tool-mcp.mjs`). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.28 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the OpenCode harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code.

The `@ai-sdk/harness-opencode` tool is an HarnessV1 adapter backed by @openai/codex-sdk, which drives the codex command line interface. Prior to version 1.0.29, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (the Codex CLI shim). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools, including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.29 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the Codex harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code.

SurrealDB before 3.1.0 evaluates user-supplied WHERE clauses in SELECT statements (and SET/MERGE/CONTENT/PATCH clauses in UPDATE, UPSERT, INSERT ON DUPLICATE KEY UPDATE, and RELATE update-variant statements) against full record data before enforcing PERMISSIONS FOR SELECT WHERE restrictions. An authenticated user — including Record and Scope users — can exploit this ordering flaw to read the full contents of any table in the database they are authenticated against, bypassing table-level permission checks. Exfiltration is most direct when scripting functions are enabled (--allow-scripting), but is also possible via SurrealQL's THROW statement and timing-based side channels without scripting. The vulnerability is confined to the attacker's current database and does not cross namespace or database isolation boundaries.

SurrealDB versions before 3.1.0 contain a field-level permission bypass vulnerability in JSON Patch operations that allows authenticated users to read protected fields. Attackers can use UPDATE PATCH with an empty from pointer in copy or move operations to duplicate all record fields, including those restricted by field-level SELECT permissions, into attacker-chosen destination fields.

SurrealDB versions before 3.1.0 contain an authentication bypass vulnerability in LIVE SELECT subscriptions where permission expressions referencing $value, $before, $after, or $event are evaluated against attacker-controlled bindings instead of actual documents. Authenticated subscribers can bind chosen values to these parameter names and register LIVE SELECT queries to receive notifications for records that SELECT permission expressions should have hidden.

SurrealDB versions before 3.1.0 contain a field-level SELECT permission bypass vulnerability in indexed COUNT fast paths. Attackers can execute COUNT queries on indexed fields with field-level SELECT restrictions to confirm or recover restricted field values through repeated guesses.

SurrealDB versions before 3.1.4 fail to properly enforce SELECT permissions on array elements (field.*) for record users, leaking denied array elements instead of hiding them. Attackers with record scope access can read array elements that element-level permissions should deny by exploiting incorrect index handling during permission filtering.

SurrealDB versions 3.1.0 before 3.1.5 fail to enforce field-level SELECT permissions when records are accessed through graph-edge or back-reference traversals. Attackers with table-level SELECT access can read field values hidden by field-level permissions by materializing records through graph traversals instead of direct table scans.

SurrealDB versions before 3.2.0 contain a permissions bypass vulnerability where data-modifying statements within PERMISSIONS clauses execute with enforcement disabled. Attackers with permission to perform a guarded operation can write to tables they lack permission for by embedding CREATE, UPDATE, DELETE, or UPSERT statements in the PERMISSIONS clause, causing unintended writes and data corruption.

IBM Verify Identity Access 11.0 through 11.0.2 and IBM Security Verify Access 10.0 through 10.0.9.1 and IBM Verify Identity Access Container 11.0 through 11.0.2 and IBM Security Verify Access Container 10.0 through 10.0.9.1 could allow an attacker with read-only privileges to make unauthorized modifications and deployments outside of their assigned permissions.

SurrealDB before 3.1.5 fail to apply field-level SELECT permissions to ORDER BY clauses, allowing authenticated users to leak the relative ordering of restricted field values. Attackers can issue ORDER BY queries on indexed restricted fields to recover the hidden values' sort order across records, even though the field itself returns null as intended.

Dendrite through 0.13.8 contains an improper access control vulnerability in the syncapi /context endpoint (syncapi/routing/context.go) that allows authenticated local users to access post-leave room state events by exploiting a flawed membership check that evaluates only the RoomExists field while ignoring IsInRoom, HasBeenInRoom, and Membership fields. Attackers who have left a room can call the rooms context API endpoint for a previously permitted event and receive unfiltered current room state that the /messages and /sync endpoints correctly withhold.