/

CVE-2026-55428

Severity:

HIGH RISK

(8.2)

Authorization Flaw (CWE-285) in Coder Tailnet Routing

CVE-2026-55428

Analyzed

Vulnerability type

Category:

Authorization and Access Control

CWE:

CWE-285, CWE-863

Impact:

Unauthorized Network Access and Traffic Hijacking

Severity:

HIGH RISK

(8.2)

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

Coder's tailnet coordinator previously accepted agent-supplied AllowedIPs without validating that these routes were tied to the authenticated agent identity.

An attacker controlling a malicious or compromised agent could advertise unauthorized prefixes, causing peers to install incorrect WireGuard routes and potentially intercept or impersonate other endpoints on the tailnet.

Affected context

This vulnerability affects Coder deployments that use the tailnet coordinator to manage remote development environments over WireGuard tunnels.

It is relevant when agents are allowed to register and update their AllowedIPs with the coordinator, and other peers automatically trust and install these routes into their local WireGuard configuration.

Environments with many agents, dynamic provisioning, or weaker controls around agent enrollment are more exposed, as a single misbehaving or compromised agent can influence routing for other peers.

Exploit preconditions

  • Attack surface: Tailnet coordinator API where agents register their identity and tunnel parameters.
  • Required access: Control over a registered or registering agent (e.g., a malicious agent, compromised agent token, or misused enrollment mechanism).
  • Trigger point: The agent supplies crafted AllowedIPs prefixes while establishing or updating its tunnel with the coordinator.
  • Why exploitation is possible: The coordinator previously validated that Addresses derived from the agent's authenticated UUID but performed no equivalent authorization check for AllowedIPs. As a result, it forwarded the agent-provided prefixes verbatim to peers, which then installed those routes into their WireGuard configuration.

Why this happens

The issue stems from an incomplete authorization check on routing configuration fields associated with an authenticated agent.

While the system correctly tied an agent's primary Addresses to its UUID, it treated AllowedIPs as trusted, agent-controlled data and did not verify that the advertised prefixes were legitimately associated with the same identity.

This maps to CWE-285: Improper Authorization and is closely related to CWE-863: Incorrect Authorization, since the coordinator enforced identity checks for one closely related field but omitted equivalent checks for another, logically linked field.

Abstractly, this is an access-control bug where policy enforcement is applied inconsistently across parameters that all influence which resources (in this case, network prefixes) an actor can affect.

Potential impact

  • An attacker-controlled agent could announce AllowedIPs that cover IP ranges belonging to other agents.
  • Peers may route traffic destined for other agents or services through the malicious agent, enabling traffic interception or modification.
  • The attacker could impersonate other development environments at the network layer, potentially accessing services or data intended for those environments.
  • Misrouted traffic could cause availability issues, such as blackholing of connections or routing loops within the tailnet.
  • Depending on surrounding network policies, the attacker may pivot to additional internal resources reachable only across the tailnet.

General mitigation guidance

Long-term mitigation focuses on consistent authorization of all identity-bound routing attributes and tightening agent trust boundaries:

  • Apply the same identity-based validation to AllowedIPs and similar routing-related fields as is used for primary addresses, ensuring each prefix is explicitly bound to the authenticated agent identity.
  • Reduce reliance on agent-supplied routing metadata by deriving acceptable prefixes from centrally managed configuration or policy rather than from the agent itself.
  • Segment tailnet participants so that agents have narrowly scoped reachable networks, and avoid designs where a single agent can influence large swaths of routing.
  • Enforce strong controls around agent provisioning, including short-lived credentials, rotation, and revocation to limit the impact of a compromised agent identity.
  • Configure underlying infrastructure and services with least-privilege network ACLs and service-level authorization, so that incorrect routing alone does not grant data or control-plane access.

How teams detect issues like this

Effective detection relies on reviewing both control-plane logs and routing configuration changes:

  • Monitor coordinator logs for agents advertising AllowedIPs prefixes that appear outside their expected ranges or patterns (for example, prefixes corresponding to other tenants, projects, or known environment segments).
  • Implement policy-based checks or admission controls that flag or reject AllowedIPs updates that do not match an agent's assigned ranges.
  • Use configuration compliance tools or scripts to periodically compare deployed WireGuard peer configurations against centrally defined routing policies.
  • During code review, look for authorization logic that validates one identity-linked field (e.g., addresses) but not related parameters (e.g., routes, tags, resource scopes).
  • Static or dynamic analysis tools can be configured to highlight flows where user- or agent-supplied configuration is propagated into network control-plane decisions without intermediate policy enforcement.

Sources

SourceLink
github.comgithub.com/coder/coder/pull/26144
github.comgithub.com/coder/coder/releases/tag/v2.29.17
github.comgithub.com/coder/coder/releases/tag/v2.32.7
github.comgithub.com/coder/coder/releases/tag/v2.33.8
github.comgithub.com/coder/coder/releases/tag/v2.34.2
github.comgithub.com/coder/coder/security/advisories/GHSA-wrq8-fcv5-8hvp

Frequently asked

questions

Exploitation typically requires control of an agent that can register with the tailnet coordinator. It is not a generic unauthenticated internet attack, but a malicious or compromised agent can abuse it remotely once enrolled.

The attacker needs the ability to run an agent that can authenticate to the coordinator—usually by possessing valid enrollment credentials or compromising an existing agent. Without this level of control, they cannot influence the AllowedIPs accepted by the coordinator.

The safest pattern is to apply consistent, identity-based authorization checks to all routing-related parameters, including AllowedIPs, and to derive permitted prefixes from centrally controlled policy tied to the agent's identity rather than trusting agent-supplied values.

Prevention involves treating all configuration that affects access—network routes, addresses, resource scopes—as subject to explicit authorization checks, designing APIs so that clients select from server-defined options rather than submitting arbitrary values, and regularly reviewing authorization logic for gaps where related fields are validated inconsistently.

Find out what's already
exploitable in your codebase.