Auth & access CVEs.
Published CVEs classified as auth & access, ranked by exploit odds.
-
CVE-2026-124577
A weakness has been identified in Intelbras SG 2404 PoE L2+ 1.00.70/1.00.72. The affected element is an unknown function of the file /EXCU_SHELL of the component Web Management Interface. Executing a manipulation can lead to improper authentication. The attack can be launched remotely. The exploit has been made available to the public and could be used for attacks. Upgrading to version 1.00.74 is sufficient to fix this issue. Upgrading the affected component is advised.
-
CVE-2026-59105
Jumis (formerly Tildes Jumis) creates and uses a Microsoft SQL Server login named ##Tilde_JumisAdmin## whose password is derived deterministically from a single non-secret input - the SQL Server name (@@servername). The derivation algorithm is hard-coded in the client application and is identical across every installation; it contains a hard-coded secret key for entropy. Because the only input (the server name) is not secret - the instance suffix is hard-coded and the host name is discoverable via the SQL Server Browser service (UDP 1434), the TDS pre-login response, or DNS/NetBIOS - any party able to reach a Jumis SQL Server can compute the password offline and authenticate. The ##Tilde_JumisAdmin## login is a member of the sysadmin server role. Successful authentication therefore grants full administrative control of the SQL Server instance and read/write access to every customer's accounting database hosted on it. The on-premise server installation also programmatically enables xp_cmdshell during routine operation, so sysadmin access escalates to operating-system command execution on the database server.
-
CVE-2026-123158
sshd in OpenSSH before 10.6p2 on Linux, when musl libc is used without PAM, lacks a workaround for a password authentication bypass (related to misinterpretation of whether a string is intended to be a password hash). For some configurations of accounts, any password between 256 and 1024 characters succeeds. An example of a platform using musl libc without PAM is Alpine Linux.
-
CVE-2026-84183
Incorrect Authorization in the updateLayout simple-method in applications/content/minilang/layout/LayoutEvents.xml in Apache OFBiz 24.09 before 24.09.08 on all platforms allows a remote authenticated attacker holding only OFBTOOLS_VIEW and CONTENTMGR_VIEW to execute arbitrary commands as the OFBiz process user, by using updateLayout's authorization check — which validates only the fixed CONTENT_CREATE operation rather than an update permission on the actual Content record and an independently selected DataResource — to relink an existing RESPONSE-purpose Content record and convert an arbitrary, independently selected DataResource into an OFBIZ_FILE/FTL resource pointed at the OFBiz access log, poisoned via a crafted User-Agent header and evaluated through the normal authenticated showContent route. Users are recommended to upgrade to version 24.09.08, which fixes this issue.
-
CVE-2026-121933
A flaw has been found in Linksys E5350 1.0.00.037. This impacts the function save_settings of the file /wizard/apply.cgi of the component Setup Wizard. Executing a manipulation of the argument wizard_router_pwd can lead to missing authentication. The attack may be launched remotely. The exploit has been published and may be used. The vendor was contacted early about this disclosure but did not respond in any way.
-
CVE-2026-118066
The Appointment Booking Plugin – LatePoint | Calendar & Scheduling for WordPress plugin for WordPress is vulnerable to Authentication Bypass via Account Takeover in all versions up to, and including, 5.7.6. This is due to the `OsModel::build_conditions_query()` function in `lib/models/model.php` comparing WHERE-clause values against the literal strings `'IS NULL'` and `'IS NOT NULL'` and concatenating a match directly into SQL without binding it as a parameter, a flaw that neither `sanitize_text_field()` nor the 5.7.5 `is_scalar()` hardening guard closes, since both pass the scalar sentinel through unchanged. This makes it possible for unauthenticated attackers to send `params[password_reset_token]=IS NOT NULL` to the public `customer_cabinet__change_password` AJAX route to match every non-null `account_nonse` value, resolving to the lowest-id customer and overwriting that customer's password to an attacker-chosen value — granting full access to their bookings, orders, and PII, and, when the `wp_users_as_customers` option is enabled, takeover of the linked WordPress subscriber or customer-role account. The attacker can first identify the target customer's email by sending `params[order_intent_key]=IS NOT NULL` to the public `steps__start_from_order_intent` route, which returns the name and email from the first matching order intent, and can then self-arm the required 15-minute password-reset freshness window by triggering a reset for that customer directly.
-
CVE-2026-82907
Authentication Bypass by Capture-replay in ExternalLoginKeysManager in Apache OFBiz 24.09 before 24.09.08 on all platforms allows a remote unauthenticated attacker who captures a live externalLoginKey value to obtain a complete, cookie-free, reusable session as the user who issued it, in every webapp that user can reach, because checkExternalLoginKey runs as a global preprocessor ahead of both the CSRF and authentication checks and accepts the key as a bearer credential that is never bound to a client, never expires, and is not invalidated after use, while the framework itself both appends live keys to every cross-webapp link it renders and, via WidgetWorker's unvalidated inter-app link target, lets an authenticated user's key be attached to an attacker-chosen absolute URL and delivered directly to the attacker's own server. Users are recommended to upgrade to version 24.09.08, which fixes this issue.
-
CVE-2026-17313
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote attacker to obtain sensitive information and take control of gateway configuration due to authentication bypass via an alternate path.
-
CVE-2026-16720
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote attacker to obtain valid access tokens for other users due to improper authentication. Authentication against a connected directory service is not sufficiently enforced, which could allow another user's identity to be assumed.
-
CVE-2026-121615
A vulnerability was determined in Linksys E5600 1.1.0.26. Affected is an unknown function of the file /cgi-bin/mesh_login.cgi of the component Login Handler. Executing a manipulation can lead to improper authentication. The attack can be launched remotely. The exploit has been publicly disclosed and may be utilized. The vendor was contacted early about this disclosure but did not respond in any way. This vulnerability only affects products that are no longer supported by the maintainer.
-
CVE-2026-107622
The BuddyForms plugin for WordPress is vulnerable to Privilege Escalation in all versions up to, and including, 2.10.5 via a form-type confusion attack through the `bf_post_type` parameter. This is due to the public frontend submission handler `buddyforms_form_response_no_ajax()` authorizing logged-out submissions based solely on the target form's `public_submit` flag while independently trusting the attacker-supplied `$_POST['bf_post_type']` to determine the post type of the newly created object — a cross-check introduced in 2.10.2 applies only to existing posts (`post_id != 0`) and is completely bypassed for new-object creation — combined with the `save_post` listener `buddyforms_edit_form_save_meta_box_data()` lacking both a `current_user_can()` capability check and a form-builder nonce, causing it to unconditionally persist attacker-supplied `$_POST['buddyforms_options']` for any newly inserted `buddyforms` CPT post. This makes it possible for unauthenticated attackers to insert a rogue BuddyForms form configuration with an attacker-controlled `new_user_role` value into the live global form registry, then submit a registration through that form to create a new WordPress account with administrator-level privileges and achieve full site takeover. Exploitation requires the WordPress site to have "Anyone can register" enabled and at least one public BuddyForms form attached to a page, both of which are standard BuddyForms deployment configurations.
-
CVE-2026-104014
The Naimur Email OTP Verification for WooCommerce plugin for WordPress is vulnerable to authentication bypass in versions up to, and including, 2.1.0. This is due to the plugin's password-reset finalization endpoint (the wp_ajax_nopriv_ceovw_update_new_password action handled by CEOVW_AJAX::update_new_password()) authorizing the reset solely by the presence of a database row marking an email as 'VERIFIED', with no timestamp check, no session binding, and no per-reset token — combined with mark_otp_as_verified_for_reset() leaving the row in that state after a user enters their code (abandoning the flow leaves the row behind), and a scheduled 'ceovw_cleanup_expired_otps_event' cron that has no add_action() handler anywhere in the plugin so cleanup never runs. This makes it possible for unauthenticated attackers to take over any account whose owner previously requested and verified a password-reset code and did not immediately finish choosing a new password — the attacker needs only the victim's email address and a nonce that the plugin itself localizes publicly into the CEOVW_APP global on the WooCommerce account page for logged-out visitors, after which they call the endpoint, overwrite the victim's password, and are returned a live wordpress_logged_in_ session cookie for that user (full site takeover if the account is an administrator).
-
CVE-2026-17278
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote attacker to gain unauthorized access to sensitive API catalog and runtime configuration data due to improper authentication. Information supplied by the client is not sufficiently verified, which could allow a caller to appear as a trusted party.
-
CVE-2026-115574
Zulip is an open-source team collaboration tool. From 11.0 until 12.2, organizations with web-public channels enabled exposed the GET /json/users endpoint to logged-out spectators, and a missing access-control check allowed include_custom_profile_fields=true to return every user's custom profile field values in addition to the documented limited directory data. This exposed those custom profile field values to unauthenticated visitors. This issue is fixed in version 12.2.
-
CVE-2026-107154
A session validation vulnerability has been identified in the SonicOS management interface, which in specific conditions could potentially enable a remote unauthenticated attacker to perform unauthorized operations.
-
CVE-2026-123024
A vulnerability has been found in agentscope-ai AgentScope up to 2.0.11. This impacts the function get_current_user_id of the file src/agentscope/app/deps.py of the component Agent Service. Such manipulation of the argument X-User-ID leads to missing authentication. The attack can be executed remotely. The exploit has been disclosed to the public and may be used. The pull request to fix this issue awaits acceptance.
-
CVE-2026-18838
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote attacker to obtain sensitive information and modify API configurations due to improper authentication. Identity information supplied in requests is not sufficiently verified, which could allow a caller to impersonate another user.
-
CVE-2026-121639
A Missing Authentication for Critical Function vulnerability in WatchGuard AuthPoint Gateway allows a remote unauthenticated attacker with network access to the AuthPoint Gateway host to disrupt authentication services and delete the local directory synchronization configuration.
-
CVE-2026-109469
Authorization bypass through user-controlled key in Entra Device Registration allows an unauthorized attacker to elevate privileges over a network.
-
CVE-2026-17158
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote authenticated attacker to gain elevated privileges due to improper authentication.
-
CVE-2026-54384
VitaFit cloud API before 2026-10-30 contained an authentication bypass vulnerability that allows remote attackers to obtain a full session for a victim's account by submitting a linked Google or Facebook account identifier to the POST /front/login/sns/ endpoint without providing proof of ownership from the identity provider. Attackers can forge a login using an identifier exposed through account enumeration to read and modify the victim's profile and delete the account's authentication record, preventing the legitimate user from logging in.
-
CVE-2026-59104
UgCS server before 6.2 contains an authentication bypass in the HCI protocol login handler (LoginHandler) exposed on TCP port 3334. When the user database contains exactly one user (the default single-operator installation), a LoginRequest with empty userLogin and userPassword fields is interpreted as \"use the only user\" rather than as invalid credentials, and the server returns a fully authenticated session. Because the service listens on all interfaces (0.0.0.0) without TLS by default, an unauthenticated remote attacker with network access to port 3334 can obtain an authenticated session without knowing any credential and then issue privileged commands, including commanding connected drones and uploading flight routes.
-
CVE-2026-123963
A security vulnerability has been detected in OpenBMB Internet of Agents up to 9d78e88e47160eb42ee1ee8538c05e1c2c47877a. The impacted element is the function register of the file im_server/app.py of the component Agent Registry. Such manipulation of the argument desc/name leads to missing authentication. The attack can be launched remotely. The exploit has been disclosed publicly and may be used. This product does not use versioning. This is why information about affected and unaffected releases are unavailable. The project was informed of the problem early through an issue report but has not responded yet.
-
CVE-2026-123655
A flaw was found in Moodle. Insufficient access control checks on H5P content files allow unauthenticated remote attackers to view restricted materials, including content stored in the system context and files belonging to hidden or unavailable course modules. This issue results in unauthorized information disclosure of protected content.
-
CVE-2026-19868
An authentication enforcement vulnerability in SonicOS SSLVPN service may allow a user with valid credentials for an alternate authentication method to bypass SAML authentication controls. Successful exploitation may also bypass SAML-enforced multi-factor authentication (MFA).
-
CVE-2026-118514
In Symfony http-foundation prior to versions 5.4.56, 6.4.50, 7.4.23, and 8.1.11, Symfony\Component\HttpFoundation\RequestMatcher, and RequestMatcher\MethodRequestMatcher which replaces it from 6.2.x on, matches a request against a list of HTTP methods. These compared the request method to the list verbatim, while the router serves a HEAD request with the GET handler, as HTTP semantics require. This came after the incomplete fix of CVE-2026-45075 https://github.com/advisories/GHSA-6439-2f28-8p8q , which addressed the same gap in the #[IsGranted], #[IsSignatureValid] and #[IsCsrfTokenValid] attributes only. Security rules configured to apply only to GET requests could therefore be bypassed by sending a HEAD request to the same URL. Although HEAD responses contain no body, attackers could still access response headers and trigger the controller's execution, including any side effects. This issue also affects firewalls configured with methods: [GET]. A HEAD request would not match that firewall and could instead be handled by another matching firewall or no firewall at all. Applications using these vulnerable request matchers directly, including through a custom firewall request_matcher, could also be affected. The issue affects symfony/http-foundation through v5.4.55, v6.x through 6.4.49, v7.x through v7.4.22, and v8.x through v8.1.10.
-
CVE-2026-17297
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote attacker to perform an account takeover due to an open redirect vulnerability. The redirect address used during sign-in is not sufficiently validated, which could allow authentication results to be sent to an attacker-controlled location.
-
CVE-2026-77439
Immich is a self-hosted photo and video management solution. Prior to 3.0.3, an authenticated user with editor access to a shared album can change the existing owner's role because album member role updates do not protect the owner from modification. By demoting the owner and then promoting themselves, an editor can take ownership of the album, remove the original owner, and delete the album. This vulnerability is fixed in version 3.0.3.
-
CVE-2026-92188
Quest Foglight for Databases ForgeMBeanServerSupport Missing Authentication Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Quest Foglight for Databases. Authentication is not required to exploit this vulnerability. The specific flaw exists within the ForgeMBeanServerSupport module. The issue results from the lack of authentication prior to allowing access to functionality. An attacker can leverage this vulnerability to execute code in the context of SYSTEM. Was ZDI-CAN-29168.
-
CVE-2026-119495
Missing Authorization vulnerability in WebberZone Better Search better-search allows Exploiting Incorrectly Configured Access Control Security Levels.This issue affects Better Search: from n/a through 4.5.2.
-
CVE-2026-17240
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote attacker to bypass security restrictions due to improper authentication. Authentication checks are not applied consistently, which could allow protected resources to be reached without valid credentials.
-
CVE-2026-95860
node-opcua is an OPC UA implementation for TypeScript and Node.js. From 2.42.0 until 2.180.0, namespace-level access control configured through setDefaultRolePermissions is not honored by the runtime enforcement path for nodes without explicit per-node RolePermissions. SessionContext.getApplicableRolePermissions reads a missing or null DefaultRolePermissions metadata value instead of the configured namespace field, and the null result is treated permissively. An anonymous network client can consequently write variables with CurrentWrite access or call bound methods when an operator configured a restrictive namespace default and relied on inheritance. Servers with no declared policy, nodes with explicit per-node permissions, and correctly bound programmatic namespace metadata are outside the authorization-bypass case. This issue is fixed in version 2.180.0.
-
CVE-2026-123023
A flaw has been found in agentscope-ai AgentScope up to 2.0.8. This affects the function check_permissions of the file src/agentscope/tool/_builtin/_bash.py of the component Bash Tool. This manipulation causes incorrect authorization. Remote exploitation of the attack is possible. The exploit has been published and may be used. Upgrading to version 2.0.9 mitigates this issue. Patch name: 6b43b13a46401e86abd3defeed9a8bfe5b37bc6d. It is recommended to upgrade the affected component.
-
CVE-2026-123032
A security flaw has been discovered in OpenBMB Internet of Agents up to 9d78e88e47160eb42ee1ee8538c05e1c2c47877a. This affects the function ConnectionManager.connect of the file im_server/app.py of the component WebSocket Connection Manager. The manipulation of the argument agent_name results in missing authentication. It is possible to launch the attack remotely. The exploit has been released to the public and may be used for attacks. This product takes the approach of rolling releases to provide continious delivery. Therefore, version details for affected and updated releases are not available. The project was informed of the problem early through an issue report but has not responded yet.
-
CVE-2026-121555
Foxel through 2.4.0 contains a missing authentication vulnerability that allows unauthenticated attackers to read files by requesting the /api/fs/stream and /api/fs/thumb endpoints. Attackers can supply a VFS mount prefix and relative path to retrieve full file contents or thumbnails across all mounts, bypassing PathRule permission grants.
-
CVE-2026-17326
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote authenticated attacker to bypass security restrictions due to improper authentication.
-
CVE-2026-17360
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote attacker to gain full administrative control due to the use of hard-coded credentials.
-
CVE-2026-17362
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote attacker to create, modify, or delete tenants due to missing authorization.
-
CVE-2026-118505
Improper Restriction of Excessive Authentication Attempts vulnerability in team-alembic ash_authentication allows an unauthenticated attacker to guess passwords and one-time passwords without limit. The password, magic link and OTP strategies enforce brute_force_strategy by adding a preparation only to the actions they generate. When the resource defines one of those actions itself, the strategy uses it as written and nothing checked that it carried the preparation, so the setting compiled and applied no limit. The generators produce exactly this: mix ash_authentication.install writes sign_in_with_password, request_password_reset_token and request_magic_link, mix ash_authentication.upgrade then adds brute_force_strategy to those strategies, and mix ash_authentication.add_strategy.otp writes its OTP actions without the preparation. Separately, with registration_enabled? true the OTP sign-in is a create action, which preparations do not run on, so brute_force_strategy {:preparation, module} never applied to it. This issue affects ash_authentication: from 5.0.0-rc.4 before 5.0.0-rc.16.
-
CVE-2026-104768
The AutomatorWP – No-Code Workflow Automation, Integration & Webhooks Plugin, now with AI plugin for WordPress is vulnerable to Privilege Escalation in all versions up to, and including, 6.0.4. This is due to missing privilege-ceiling validation in the `create_user()` function within the anonymous-user action (and equally in the Create User and Add/Change Role WordPress-integration actions), which verify only that the submitted role exists via `automatorwp_get_editable_roles()` without confirming the role falls at or below the acting context's privilege level — allowing attacker-controlled trigger tag data (e.g., a `{comment_author}` tag bound to the Role option) to supply any role, including `administrator`, unchallenged. This makes it possible for unauthenticated attackers to create a new administrator-role account tied to an attacker-controlled email address, receiving the generated password via WordPress's new-user notification email and thereby gaining full administrative access to the site. Exploitation requires the site to have an anonymous automation configured with the Role option bound to a dynamic trigger tag such as `{comment_author}`, and is triggered when a site administrator performs the routine action of approving a comment — an action that provides no in-UI indication that it will mint a new privileged account from the comment author field.
-
CVE-2026-17172
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote authenticated attacker to bypass security restrictions due to improper authentication. Information supplied by the client is not sufficiently verified, which could allow a caller to impersonate another party.
-
CVE-2026-17066
IBM API Connect 12.1.0.0 through 12.1.1.2 could allow a remote authenticated attacker to obtain sensitive information and modify or delete data belonging to other tenants due to missing authorization.
-
CVE-2026-16716
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote authenticated attacker to bypass security restrictions to access and modify sensitive information due to improper restriction of security token assignment.
-
CVE-2026-123550
OpenAM before 16.1.4 contains an authentication bypass vulnerability in its SAML v2.0 service provider that accepts unsigned assertions on the ECP (PAOS) and HTTP-Artifact bindings when WantAssertionsSigned is false. Unauthenticated remote attackers can submit forged assertions to obtain sessions as any mapped user, including administrators, and hijack single sign-on to dependent applications.
-
CVE-2026-58273
Rucio is a scientific data management framework. From 1.16.1 until 35.9.0, 38.6.0, and 41.0.0rc1, the /credentials/signurl endpoint reads X-Rucio-Account instead of the account stored in request.environ['issuer'] by token authentication. Any authenticated low-privilege user can supply root as the header value and pass the root-only get_signed_url permission check. The server then returns a presigned read, write, or delete URL for an attacker-selected object using stored S3, GCS, or Swift credentials, and X-Rucio-VO can select another tenant in multi-VO deployments. This permits data disclosure, modification, deletion, and cross-tenant access on cloud-backed RSEs. This issue is fixed in versions 35.9.0, 38.6.0, and 41.0.0rc1.
-
CVE-2026-120554
The Import and export users and customers plugin for WordPress is vulnerable to Privilege Escalation in all versions up to, and including, 2.5.11 This is due to a non-atomic multi-option write sequence in the `ACUI_Cron::save_settings()` function, which persists `acui_cron_activated`, `acui_cron_configured_by`, `acui_cron_user_id`, and `acui_cron_path_to_file` before reaching the authorization gate that calls `wp_die()` when a non-promoter submits a role. This makes it possible for authenticated attackers holding a custom delegated role with `create_users` but without `promote_users` or `edit_users` to escalate their privileges to Administrator, provided a prior administrator cron-tab save exists and no successful save intervenes before the next unattended cron tick processes the attacker's repointed CSV file.
-
CVE-2026-17280
IBM API Connect 10.0.8.0 through 10.0.8.12 and 12.1.0.0 through 12.1.1.2 could allow a remote attacker to perform unauthenticated brute-force attacks to obtain valid credentials due to improper restriction of excessive authentication attempts.
-
CVE-2026-121939
Databasement before 1.10.9 contains an authentication bypass vulnerability in OAuthService::findOrCreateLocalUser() that links OIDC identities to local users by email without checking email_verified. Attackers can register an identity provider account with a victim's unverified email to log in as any user, including the super-admin, and retain access after password resets.
-
CVE-2026-118267
The WP Data Access – App Builder for Tables, Forms, Dashboards, Charts & Maps plugin for WordPress is vulnerable to Incorrect Authorization leading to Privilege Escalation in all versions up to, and including, 5.5.89 This is due to the plugin's write endpoints (`wpda/app/update`, `wpda/app/delete`, `wpda/app/insert`) passing the attacker-supplied `key` and `val` parameters directly to `$wpdadb->update`, `$wpdadb->delete`, and `$wpdadb->insert` without enforcing the container's row-filter (`defaultWhere`) or per-column allow-list (`isSelected`) restrictions that are correctly applied on the read endpoints. This makes it possible for authenticated attackers, with Subscriber-level access and above, to write to arbitrary rows and columns outside the configured restrictions — including overwriting the `user_pass` field of the WordPress administrator account in `wp_users` — and subsequently authenticate as the administrator, achieving full site takeover. Exploitation requires an administrator to have published a Data App that enables write transactions (update, delete, or insert) for Subscriber-level or anonymous users, which is an explicitly advertised use-case of the plugin. Exploitation by unauthenticated attackers requires that the administrator has published a Data App whose rest_api.authorization is 'anonymous' and whose container has transactions.update (or delete/insert) enabled. On a default installation, where that is not configured, the vulnerability requires attackers with subscriber-level access and above.
-
CVE-2026-111771
The Model Context Protocol TypeScript SDK provides TypeScript client and server implementations of the Model Context Protocol. From 1.6.0 until 1.32.0 in @modelcontextprotocol/sdk, from 2.0.0 until 2.3.0 in @modelcontextprotocol/server, from 2.0.0 until 2.0.2 in @modelcontextprotocol/express, and from 2.0.0 until 2.3.1 in @modelcontextprotocol/server-legacy, server-side bearer authentication could accept a valid access token issued for a different service because the SDK did not enforce the token audience. The affected surface is server-side HTTP bearer authentication. The trigger is a valid token whose audience names another service. The authorization server must issue tokens for multiple services, and the application's verifyAccessToken() must not independently validate aud. Applications already validating aud inside verifyAccessToken() are not affected. An attacker with a token for another service could authenticate to the MCP server and obtain the confidentiality and integrity capabilities granted to an authenticated caller. Upgrading alone leaves the new audience validation disabled. Operators must enable audience validation and provide the token audience together. Enabling only the expected audience check causes every request to receive HTTP 401. This issue is fixed in @modelcontextprotocol/sdk 1.32.0, @modelcontextprotocol/server 2.3.0, @modelcontextprotocol/express 2.0.2, and @modelcontextprotocol/server-legacy 2.3.1.
A list of CVEs whose flaw type is auth & access, sorted by how likely they are to be exploited (EPSS). Classification uses the CWE weakness id where present, and a small local model otherwise.