Summary
Avo: Direct attachment upload endpoint lacks upload authorization and bypasses field-level upload policy
Full technical description
Avo's direct attachment upload endpoint lacks server-side upload authorization and bypasses the documented field-level upload policy methods such as upload_{FIELD_ID}?.
An authenticated Avo user who can reach the Avo attachment upload endpoint can replace or add attachment content, including binary content, filename, and content-type metadata, on a resolved record even when both update? and upload_<field>? policies deny the operation.
This primarily affects multi-role Avo Pro/Advanced-style deployments where non-administrator or restricted operator users can reach Avo and per-record or per-field operations are expected to be enforced by policies.
Details
Avo exposes a direct attachment upload endpoint:
POST /<avo-root>/avo_api/resources/:resource_name/:id/attachments/
The vulnerable code path is Avo::AttachmentsController#create:
app/controllers/avo/attachments_controller.rb:9-24
Current behavior:
def create
blob = ActiveStorage::Blob.create_and_upload! io: params[:file].to_io, filename: params[:filename]
association_name = BaseResource.valid_attachment_name(@record, params[:attachment_key])
if association_name
@record.send(association_name).attach blob
elsif params[:key].blank?
raise ActionController::BadRequest.new(...)
end
render json: {
url: main_app.url_for(blob),
href: main_app.url_for(blob)
}
end
This endpoint creates an ActiveStorage::Blob before validating the requested attachment association and before any upload authorization could be enforced. If attachment_key resolves to an association, the blob is attached to the target record. If the request uses the key-based/Trix path, the endpoint can still persist the blob and return url/href even when no attachment association is resolved on the target record.
The controller never calls:
@resource.authorization.authorize_action("upload_<field>?", record: @record, ...)@resource.authorization.authorize_action("update?", record: @record, ...)
This is inconsistent with the rest of Avo's file authorization implementation. The field-level file authorization concern defines upload authorization as:
lib/avo/fields/concerns/file_authorization.rb:11-12lib/avo/fields/concerns/file_authorization.rb:25-27
def can_upload_file?
authorize_file_action(:upload)
end
def authorize_file_action(action)
authorize_action("#{action}_#{id}?", record: record, raise_exception: false)
end
That upload policy is used by UI components to decide whether to render upload controls, but the server-side upload endpoint does not enforce the same policy. A user can therefore bypass the policy by directly POSTing to the endpoint.
By contrast, attachment deletion does call attachment-specific authorization:
app/controllers/avo/attachments_controller.rb:27-65
def destroy
if authorized_to :delete
...
end
end
def authorized_to(action)
@resource.authorization.authorize_action("#{action}_#{params[:attachment_name]}?", record: @record, raise_exception: false)
end
The asymmetry is:
destroy: callsdelete_<attachment_name>?create: does not callupload_<attachment_key>?orupdate?
The field-level upload authorization intent is also documented by Avo:
- Issue #1624 requested the "Ability to police each file upload/download/delete".
https://github.com/avo-hq/avo/issues/1624 - PR #1625 introduced
upload_cover_photo?as an example upload policy method.
https://github.com/avo-hq/avo/pull/1625 - PR #1667 clarified the policy semantics: "From now on, only field level
authorization will be considered. If it is defined and returns true, the
action will be granted; otherwise, it will not."
https://github.com/avo-hq/avo/pull/1667 - The current Avo authorization documentation lists
upload_{FIELD_ID}?as the
policy method controlling whether a user can upload an attachment, stating:
"Controls whether the user can upload the attachment."
https://docs.avohq.io/4.0/authorization.html - The same documentation says the same
upload_file?policy method will be used
to "authorize the file upload" in action file fields.
https://docs.avohq.io/4.0/authorization.html
Affected versions observed:
- PoC-confirmed: Avo
3.31.2, commit46aa6b3bc9e3283110c39e58cfec8bb95adc1897 - Same vulnerable code path by source inspection:
origin/mainHEAD as of
2026-05-29,9e23ddc88f2b1e762e4a5ec35a6f86370ac16c73 - Same vulnerable code path by source inspection: Avo
v4.0.0.beta.26,6d339595a27f8779cb99b4aa38ddc97cb702b30f - Same vulnerable code path by source inspection:
origin/4-devat version4.0.0.beta.40,7ab9794f8b044b11b9677cdc57547d99cf96c3f3
Suggested affected range for the GitHub Security Advisory form:
>= 2.28.0
Rationale:
- The direct attachment upload endpoint exists without upload authorization from
commitab5f5970e2aa76e6ca0a95bf04f510ba7ed5e858(feature: trix attachments, 2021-04-24), first included in tagv1.4.0. - The documented field-level upload policy bypass is confirmed from commit
667049ceeda838394214489693d088708d9da77d(feature: field level file authorization, 2023-03-12), first included in tagv2.28.0. - PR #1667 later clarified the field-level-only grant/deny semantics in commit
31b6ef94f8cc4c340a2a75eec36c838cda933ce7, first included in tagv2.30.1.
From a narrow missing-authorization perspective, the issue exists fromv1.4.0; the >= 2.28.0 range is a conservative submission range anchored to
the documented field-level upload policy boundary.
Fixed version: to be determined by maintainers.
PoC
A local request spec was used to reproduce the issue. The PoC adds pundit to
the test group and replacesAvo::Services::AuthorizationService with a test double. The test double
records every authorize_action invocation and, when invoked, delegates the
decision to a PostPolicy resolved via Pundit.policy!.
The policy setup is:
PostPolicy#update?returnsfalsePostPolicy#upload_attachments?returnsfalsePostPolicy#upload_cover_photo?returnsfalsePostPolicy#method_missingreturnstruefor all other policy methods ending
in?, simulating a user with general read access but explicit update/upload
denial- the replacement authorization service records all
authorize_actioncalls
The open-source repository does not include Avo Pro's authorization client. The
critical observation is independent of which client is plugged in: the vulnerable
endpoint never invokes authorize_action at all, so the call list remains empty.
The PoC user has roles: {admin: true}, which is required only to pass the
dummy app's coarse route-level guard:
authenticate :user, ->(user) { user.is_admin? }
The vulnerability under test is one layer below that guard: the fine-grained
record/field authorization that Avo Pro/Pundit deployments would normally
enforce. The dummy route gate is not part of Avo's upload policy decision. In a
deployment where non-admin operators reach Avo through a different
authentication rule, AttachmentsController#create would still skipauthorize_action for the upload.
Policy scoping is orthogonal to this finding. Even if apply_policy filtered
the record set during lookup, AttachmentsController#create would still not
authorize the upload action itself for records that survive the scope.
The spec demonstrates:
- Normal record update is denied by policy and does not persist changes.
- Direct upload with
attachment_key=cover_photosucceeds even thoughPostPolicy#upload_cover_photo?returnsfalse, replacing the existinghas_one_attachedcover_photoblob. - Direct upload with
attachment_key=attachmentssucceeds even thoughPostPolicy#upload_attachments?returnsfalse, adding to ahas_many_attachedassociation. - Direct upload with
params[:key]and no attachment association succeeds,
creates anActiveStorage::Blob, and returnsurl/hrefwithout attaching
the blob to the record. - The direct upload requests do not invoke the replacement authorization
service at all.
The full request-spec patch can be provided in this advisory thread if useful.
Verification environment:
- Ruby
3.3.1 - Avo
3.31.2 - Commit
46aa6b3bc9e3283110c39e58cfec8bb95adc1897
Impact
This is a missing server-side authorization vulnerability in Avo's direct
attachment upload endpoint.
The primary affected deployments are Avo Pro/Advanced-style multi-role
deployments where non-administrator or restricted operator users can reach Avo,
and per-record/per-field operations are expected to be enforced by policies.
In such deployments, an authenticated Avo user can add or replace attachment
content on a resolved record even when the host application policy denies both:
- record update, for example
update? - field-level upload, for example
upload_cover_photo?orupload_attachments?
For has_one_attached fields, the demonstrated impact is replacement of a
protected attachment field with attacker-controlled content. The attacker
controls the uploaded binary content, filename, and content-type metadata for
the reachable field.
For key-based/Trix upload flows, the endpoint can persist a blob and return a
URL even when no attachment association is resolved. This means the endpoint
should not be left as an unauthenticated blob creation and URL return path for
authenticated Avo users.
In admin-only deployments following the Avo Community pattern, practical
exposure is much lower because the only users who can reach the endpoint are
already trusted to perform record updates through the normal CRUD path.
Suggested fix direction:
- validate the requested
attachment_keybefore any blob is stored - perform upload authorization before
ActiveStorage::Blob.create_and_upload! - derive the policy method from the validated association name rather than raw
request input - apply an equivalent authorization decision to supported
params[:key]/ Trix
upload flows before blob creation or URL return - return a JSON-compatible
403 Forbiddenresponse when upload authorization is
denied
The default Avo::Services::AuthorizationService#authorize_action returnstrue in lib/avo/services/authorization_service.rb:34-35, so applications
without a custom authorization client should not see a behavior change from
adding an authorization call. Deployments with custom/Pro authorization clients
that explicitly deny upload would gain enforcement at this endpoint.
No official Avo-level workaround was confirmed in this review. Until a fix is
available, applications can reduce exposure by ensuring only fully trusted
administrators can access Avo routes.
Applications needing an immediate mitigation may override or monkey-patchAvo::AttachmentsController#create to authorize uploads before blob creation.
Any mitigation should be tested against the application's Avo authorization
client and upload UI, because the endpoint is used by Trix/file upload flows.
If a mitigation falls back to update? for key-based/Trix uploads, that is a
conservative behavior change for applications that currently allow read-only
operators to use those uploads; those deployments should replace the fallback
with an explicit rich-text upload policy.
The application does not perform an authorization check before performing a sensitive operation. Typical impact: unauthorized access to restricted functionality or data.
CVE-2026-53769 has a CVSS score of 6.5 (Medium). The vector is network-reachable, low privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (3.32.0); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
Kodem Kai can prioritize this vulnerability in your dependency tree and generate a fix recommendation.
Frequently Asked Questions
- What is CVE-2026-53769? CVE-2026-53769 is a medium-severity missing authorization vulnerability in avo (rubygems), affecting versions >= 2.28.0, < 3.32.0. It is fixed in 3.32.0. The application does not perform an authorization check before performing a sensitive operation.
- How severe is CVE-2026-53769? CVE-2026-53769 has a CVSS score of 6.5 (Medium). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which versions of avo are affected by CVE-2026-53769? avo (rubygems) versions >= 2.28.0, < 3.32.0 is affected.
- Is there a fix for CVE-2026-53769? Yes. CVE-2026-53769 is fixed in 3.32.0. Upgrade to this version or later.
- Is CVE-2026-53769 exploitable, and should I be worried? Whether CVE-2026-53769 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-53769 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-53769? Upgrade
avoto 3.32.0 or later.