/

CVE-2026-28474

Severity:

CRITICAL RISK

(9.8)

Improper Authorization in OpenClaw Nextcloud Talk Plugin

CVE-2026-28474

Analyzed

Vulnerability type

Category:

Improper Authorization

CWE:

CWE-863

Impact:

Bypass of Access Controls and Unauthorized Conversation Access

Severity:

CRITICAL RISK

(9.8)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

OpenClaw's Nextcloud Talk plugin versions prior to 2026.2.6 implement DM and room allowlists by comparing the mutable actor.name display name field rather than a stable user identifier.

By changing their Nextcloud display name to match an allowlisted user ID, an attacker can bypass access controls and join restricted conversations without proper authorization.

Affected context

This vulnerability affects deployments using OpenClaw's Nextcloud Talk plugin for access control over direct messages (DMs) and rooms.

The issue arises when the plugin relies on equality comparison of the actor.name (display name) field for allowlist checks instead of using immutable, trusted user identifiers.

Environments where users can freely modify their display names, and where allowlists are assumed to map to specific user identities, are particularly exposed.

Exploit preconditions

  • Attack surface: Nextcloud Talk conversations (DMs and rooms) protected by the OpenClaw plugin's allowlist mechanism.
  • Required access: Ability to authenticate to the Nextcloud instance with any user account and change that account's display name.
  • Trigger point: Joining or initiating a DM/room where the plugin checks access by comparing the joining user's actor.name to entries in the allowlist.
  • Why exploitation is possible: The authorization logic treats an easily spoofed, user-controlled display name as the identity key, allowing an attacker to impersonate an allowlisted identity without possessing that user's account or credentials.

Why this happens

The core issue is an authorization decision based on an untrusted, mutable attribute rather than a stable, authoritative identity.

The plugin compares the actor.name display name field directly against allowlist entries; since this field is user-controlled and not guaranteed to be unique or immutable, it does not reliably represent the actual account associated with the allowlisted user.

This maps to CWE-863: Incorrect Authorization, where the system incorrectly determines if an actor is allowed to access a resource by using an inappropriate attribute as the identity or authorization key.

At a higher level, this is a design flaw in the access control model: the plugin fails to separate presentation attributes (display names) from security attributes (user IDs or verified identifiers).

Potential impact

  • Unauthorized access to DMs that are intended for specific individuals or small groups
  • Unauthorized entry into restricted rooms and group conversations
  • Exposure of sensitive discussion content, including confidential information or internal communications
  • Impersonation in conversations, leading to social engineering or misinformation within trusted channels
  • Degradation of trust in the chat platform's access controls and audit trails

General mitigation guidance

Mitigation generally focuses on binding authorization decisions to stable, trusted identifiers and reducing reliance on user-controlled display attributes.

Recommended design patterns include:

  • Use canonical, immutable user identifiers (e.g., internal user IDs, UUIDs, or verified account identifiers) as the basis for allowlist and access checks.
  • Treat display names (actor.name or equivalent) purely as presentation attributes, never as security or identity keys.
  • Normalize and centralize authorization logic so that all DM and room access checks reference the same authoritative identity source.
  • Incorporate defense-in-depth by validating that allowlist entries correspond to existing, active accounts and by rejecting duplicate or ambiguous identifiers at configuration time.
  • Apply the principle of least privilege by minimizing default access and requiring explicit, identity-bound allowlisting for sensitive rooms.

How teams detect issues like this

Detection efforts can focus on both code-level patterns and runtime signals:

  • Static analysis and code review:
  • Flag authorization checks that rely on mutable or user-controlled fields such as display names, profile names, or email text fields instead of stable IDs.
  • Look for equality comparisons between allowlist entries and presentation attributes (name, displayName, nickname) in access control paths.
  • Review authorization modules to verify that they use canonical identifiers from the authentication subsystem rather than repeated parsing of user-facing fields.
  • Configuration and policy checks:
  • Enforce policies requiring that all access control lists reference system-level user IDs or verified identifiers.
  • Periodically audit allowlist configurations for non-unique or ambiguous entries that could be mimicked (e.g., plain strings with no associated account ID).
  • Runtime monitoring:
  • Log changes to user display names together with user IDs, and monitor for patterns where display names frequently match allowlisted identifiers shortly before room joins.
  • Alert on repeated display name changes followed by attempts to access restricted rooms, indicating potential abuse of name-based authorization.

Automated code review or SAST tools can assist by scanning for authorization logic that uses user-facing fields in security decisions.

Sources

SourceLink
github.comgithub.com/openclaw/openclaw/commit/6b4b6049b47c3329a7014509594647826669892d
github.comgithub.com/openclaw/openclaw/security/advisories/GHSA-r5h9-vjqc-hq3r
vulncheck.comvulncheck.com/advisories/openclaw-nextcloud-talk-allowlist-bypass-via-actorname-display-name-spoofing

Frequently asked

questions

Any authenticated user who can change their display name on a Nextcloud instance using the vulnerable plugin can potentially exploit this issue to access conversations protected by name-based allowlists. The impact depends on how widely the plugin is used to enforce restrictions on DMs and rooms.

No. The attacker only needs a regular Nextcloud account with permission to modify their own display name. They do not need control over the allowlist configuration or any elevated administrative role.

Rework the authorization logic so that allowlists are defined and enforced using stable, canonical user identifiers (such as internal user IDs or verified account IDs) rather than display names. Ensure that all access control checks retrieve identity from a trusted authentication source instead of user-controlled profile fields.

Establish a clear separation between identity attributes used for security (IDs, tokens, verified emails) and user-facing display attributes (names, nicknames). During design and code review, require that all authorization decisions reference canonical identity fields, apply least-privilege defaults, and use automated checks to flag any access control that relies on mutable, user-controlled data.

Find out what's already
exploitable in your codebase.