Parse Server Auth Bypass (CVE-2026-116163, CVSS 9.1): If Your Custom Login Adapter Extends AuthAdapter, It May Never Have Run

Check your own app — free, in about a minute.

The free mini-scan checks your live homepage for TLS/HTTPS basics, response security headers, cookie flags, exposed sensitive files, Firebase/Supabase backend rules, and the WordPress plugin versions your site exposes. You get a score and the finding titles — no account, no card.

Run the free mini-scan →

Parse Server Auth Bypass (CVE-2026-116163, CVSS 9.1): If Your Custom Login Adapter Extends AuthAdapter, It May Never Have Run

The most dangerous authentication bug is not broken validation code. It is validation code that looks right, passes review — and is never called. That is what happened in Parse Server: for one specific way of writing a custom login adapter, the framework silently fell back to a default that accepts any authentication data. If your app uses that pattern, logins through that provider were unauthenticated in practice, no matter how careful the code in your adapter looks.

What Happened

The flaw is tracked as CVE-2026-116163 (CWE-287, improper authentication), covered by vendor advisory GHSA-rq3w-5c6f-p7wr in the parse-community/parse-server repository.

How It Works, Plainly

Parse Server lets you add custom authentication providers — your own login method, an extra factor, a company SSO bridge — by writing a custom auth adapter. The documented way to build one is to extend the exported AuthAdapter class and implement three methods: validateSetUp, validateLogin, and validateUpdate.

The bug: for adapters built that way, those methods were never called. Parse Server used a default inherited from the base class instead — and that default accepts any authData presented to it.

In practical terms, for an affected provider an attacker could sign up, link an identity, or log in as an existing user without valid credentials — and could skip a custom adapter that was supposed to implement an additional authentication factor entirely. Your adapter code could be perfect. It simply never ran.

The patch changes the logic so inherited, unchanged methods are no longer treated as implemented: a subclass's own validation methods now run, and a subclass that implements only some of the three validate methods is rejected as not configured instead of silently accepting anything.

The Scope, Stated Straight

This is the part most write-ups will bury, so here it is plainly:

Only apps that configure a custom authentication adapter built as a subclass of AuthAdapter are affected.

If you do not use a subclass-based custom adapter, this CVE does not apply to you. Do not panic — check which kind you wrote. If you do use one, treat this as urgent, and treat any account created or linked through that provider as untrusted until you have reviewed it.

Am I Affected?

This check is server-side only, and there is an honest reason for that: Parse Server does not disclose its version to unauthenticated remote requests, and whether a deployment uses an AuthAdapter subclass is code and configuration that is never visible from outside. No outside scanner — including ours — can tell you from a URL whether you are affected. Anyone who claims otherwise is guessing.

Check it yourself, where the code lives:

npm ls parse-server

Then ask two questions:

  1. Is the version in the affected range — below 8.6.102, or on the 9.x line below 9.10.5?
  2. Does your codebase define a custom auth adapter that extends AuthAdapter (rather than a plain object)?

Yes to both means act today. Yes to the version question only means update anyway — you are one refactor away from the vulnerable pattern.

What To Do Now

  1. Update Parse Server to 8.6.102 or later (8.x) or 9.10.5 or later (9.x).
  2. If you cannot update immediately, use a vendor workaround: implement all validation in an overridden validateAuthData method that does not rely on the adapter instance's own state — or define the custom adapter as a plain object with the validation methods instead of an AuthAdapter subclass.
  3. Audit the blast radius. Review accounts that were created, linked, or logged in through the affected provider while you were on a vulnerable version. Force re-verification where it matters, and revoke sessions you cannot account for.
  4. Add a test that would have caught this. A login attempt with deliberately invalid authData for your custom provider should fail. If it succeeds, your validation is not running — whatever the code looks like.

The Bigger Lesson for AI-Built Apps

Frameworks fail silently at exactly the seams AI scaffolding writes fastest: custom adapters, guards, and middleware. The generated code follows the documentation, the class compiles, the methods are all there — and "the code is there" turns out not to be "the code runs." The defence is boring and effective: for every security control you did not write end-to-end yourself, keep one test that proves it rejects something. A control you have never seen say no may never have been asked.

That is also why we split our own reporting honestly. Exposure that is visible from outside — TLS, security headers, cookie flags, exposed files, open backend rules — is what a remote check can prove, and it is what our scan reports.

Run the free mini-scan at preshipit.com/free-scan — paste your URL and see the exposures that are visible from outside in about a minute, with a shareable result page. The full scan ($29, one-time) goes deeper: active probing, backend-rule checks, and copy-paste fixes for everything it finds. For this CVE specifically, the version check above — in your own codebase — is the one that matters.

The PreShipIt Brief — a short weekly email for people shipping vibe-coded apps: the week's important vulnerabilities and launch-safety checks, in plain English.

New and free. Double opt-in: we email you a confirmation link before anything else arrives. Unsubscribe any time.

Check your own app — free, in about a minute.

The free mini-scan checks your live homepage for TLS/HTTPS basics, response security headers, cookie flags, exposed sensitive files, Firebase/Supabase backend rules, and the WordPress plugin versions your site exposes. You get a score and the finding titles — no account, no card.

Run the free mini-scan →

← Back home · Run a free mini-scan · Scan my app — $29 · See a sample report