Severity:

HIGH RISK

(8.1)

Missing Authorization (CWE-862) in Yarbo Cloud Fleet Control

CVE-2026-7368

Deferred

Vulnerability type

Category:

Missing Authorization

CWE:

CWE-862

Impact:

Unauthorized Remote Control and Data Access

Severity:

HIGH RISK

(8.1)

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

Yarbo's cloud platform lacks proper per-device and per-user authorization, allowing any client with valid credentials to interact with all robots in the fleet.

By subscribing to wildcard topics and publishing commands using only exposed serial numbers, an attacker who compromises a single account can control or monitor arbitrary robots across the platform.

Affected context

This vulnerability affects Yarbo's cloud-based messaging and control infrastructure for connected robots.

It is most relevant in deployments where:

  • MQTT or similar pub/sub messaging is used for robot telemetry and command channels
  • Authorization checks are performed only at the account or application level, not per device
  • Topic structures allow wildcard subscriptions and per-robot command topics addressable via serial numbers

In such environments, possession of any valid cloud credential can implicitly grant broad access to robot telemetry and command interfaces.

Exploit preconditions

  • Attack surface: Cloud-based messaging endpoints (e.g., MQTT broker or equivalent pub/sub service).
  • Required access: Any valid Yarbo cloud credential (including a compromised user account or previously shared hard-coded credential).
  • Trigger point: Establishing a client connection and:
  • Subscribing to wildcard topics that aggregate telemetry from multiple robots, and/or
  • Publishing commands to per-robot topics derived from known or observed serial numbers.
  • Why exploitation is possible: The platform does not enforce device-level authorization or ownership checks on topic subscriptions and publications, so a credential is treated as globally trusted regardless of which robots it should legitimately control.

Why this happens

The underlying issue aligns with CWE-862: Missing Authorization.

Authorization decisions are made at too coarse a level (e.g., "any authenticated client") without enforcing fine-grained access control for individual devices, topics, or users.

The system implicitly trusts any authenticated client to:

  • Subscribe to wildcard topics exposing telemetry from all robots, and
  • Publish to command topics for any robot identified only by a serial number, which is available in the telemetry stream.

This represents an architectural access-control failure where the platform lacks binding between user identity, device ownership, and allowed messaging operations.

Potential impact

  • An attacker with one compromised credential could remotely control arbitrary robots across the fleet.
  • Telemetry and location-like data from all robots may be exposed to unauthorized parties.
  • Attackers could disrupt service by issuing conflicting or malicious commands to multiple robots.
  • Fleet-wide privacy violations could occur due to unauthorized monitoring of robot activity.
  • Brand damage and regulatory concerns may arise from systemic access-control weaknesses.

General mitigation guidance

Effective mitigation focuses on implementing fine-grained authorization and minimizing trust in any single credential.

Common defensive patterns include:

  • Enforcing per-device authorization, binding each robot to specific user accounts or service identities and validating that relationship on every subscribe/publish operation.
  • Designing topic hierarchies so that wildcard subscriptions are restricted or disallowed for regular clients, especially across different owners or tenants.
  • Issuing unique, revocable credentials or certificates per device and per user, with scoping that limits which topics each principal can access.
  • Applying least privilege for all messaging permissions, ensuring that default roles cannot access fleet-wide telemetry or control.
  • Separating telemetry ingestion roles from control roles, so that read access to telemetry does not automatically imply command capabilities.
  • Implementing server-side authorization policies that are centrally managed and evaluated, rather than relying on client behavior or application logic alone.

How teams detect issues like this

Organizations can detect and monitor for similar authorization weaknesses through:

  • Architectural and code review: Verifying that authorization checks exist for each operation (subscribe/publish) and that they consider user identity, device identity, and ownership relationships.
  • Static analysis and policy linting: Searching for messaging access patterns that lack explicit authorization logic or rely solely on authentication outcomes.
  • Broker-level configuration review: Auditing topic ACLs, wildcard subscription rules, and role definitions to ensure they enforce tenant and device isolation.
  • Telemetry and logging analysis: Monitoring for anomalous access patterns, such as a single credential accessing large numbers of devices or subscribing to broad wildcard topics.
  • Security testing and red teaming: Performing controlled attempts to access devices outside one account's ownership to validate enforcement of access-control boundaries.

Sources

SourceLink
github.comgithub.com/cisagov/CSAF/blob/develop/csaf_files/OT/white/2026/icsa-26-162-01.json
cisa.govcisa.gov/news-events/ics-advisories/icsa-26-162-01

Frequently asked

questions

Yes. Any attacker who obtains valid Yarbo cloud credentials can exploit this remotely via the cloud messaging interface without physical access to the robots.

The attacker needs only one valid account or credential (for example, a compromised user account or leaked application credential). No special privileges are required if authorization is missing or weak at the device level.

Implement strict, fine-grained authorization that binds each device to specific users or service identities, enforce topic-level access control for both subscriptions and publications, and prohibit or tightly scope wildcard topic access for client roles.

Adopt a security-by-design approach for multi-tenant systems, requiring explicit authorization checks for every action, defining clear access-control models (RBAC/ABAC), regularly reviewing broker and API permissions, and integrating authorization testing into CI/CD and security review processes.

Find out what's already
exploitable in your codebase.