What Is Incorrect Authorization? CWE-863 Explained
Incorrect authorization (CWE-863) lets a logged-in user do things their role forbids. How these access check flaws happen and how to limit the damage.
Incorrect authorization (CWE-863) is a flaw where an application checks whether a user may perform an action, but the check gets the answer wrong. The user is properly logged in, yet the server lets them do something their role should forbid, such as reading another tenant's files or creating administrators. It is a common root cause of privilege escalation in web applications and appliances, including the IBM Aspera Faspex CVE-2026-123218 bypass that CISA added to its exploited list in November 2026.
What is incorrect authorization?
Incorrect authorization is a software weakness where the application performs an access check but the check is wrong, incomplete or applied to the wrong data. The result is that a user can perform actions or reach data outside their permissions.
MITRE tracks it as CWE-863. It sits next to CWE-862, missing authorization, where no check exists at all. In CWE-863 the developer tried to enforce access control, and the attempt has a gap.
How is authorization different from authentication?
Authentication answers "who are you?" and authorization answers "what are you allowed to do?". An authorization flaw assumes the user is already identified and then grants them more than their role allows.
This distinction matters for triage. A bug that needs a valid login sounds less urgent than one that needs nothing. On a server that accepts external users, partners or self-registration, a valid login is easy to get, and an authorization flaw becomes the step from ordinary user to administrator.
Common ways authorization checks go wrong
Most CWE-863 bugs fall into a few patterns:
- Trusting client input. The server reads a role, user ID or account ID from the request and believes it, instead of looking it up from the session.
- Checking the wrong object. The code confirms the user may read package A, then serves package B because the ID in a later parameter was never checked.
- Inconsistent enforcement. The web interface hides an admin button, but the API endpoint behind it does not check the caller's role. Anyone who sends the request directly gets the admin action.
- Alternate message paths. A second protocol, internal API or message type reaches the same function without passing through the main access check.
- Logic errors in role comparisons. A check that should require "admin" accepts "admin or any role containing admin", or treats an empty role as allowed.
IBM described CVE-2026-123218 as an authorized user crafting "a specific message" to bypass authorization and gain full administrative privileges. IBM has not published the details, so it is not possible to say which of these patterns applies.
Why do authorization bugs survive testing?
Authorization bugs survive because functional tests use the intended role for each action, so every test passes. The bug only appears when a low-privilege user tries an action meant for someone else, and few test suites try every action as every role.
Scanners have the same blind spot. They can find injection flaws by looking at responses, but they do not know which user should be allowed to do what, so they cannot tell a correct response from an unauthorized one.
How severe is an incorrect authorization flaw?
Severity depends on what the bypass reaches. A flaw that lets a user read another user's profile is moderate. A flaw that grants full administration over a file exchange server, as in the Faspex case, is critical because the attacker gains every stored file, every account and the server's configuration.
The CVSS vector can be confusing here. IBM scored CVE-2026-123218 at 9.8 with "privileges required: none", while the description says the attacker is an authorized user. When the score and the text disagree, plan for the worse reading.
How to reduce the risk in your environment
You cannot fix a vendor's access control code, but you can limit what a flaw like this exposes:
- Shrink the user population. Disable self-registration where the business does not need it, expire guest and partner accounts, and remove users who have not logged in for months.
- Restrict the admin interface. If the product lets you limit admin functions to internal addresses or a separate listener, do it. A bypass that needs the admin endpoint fails if the attacker cannot reach it.
- Alert on new administrators. A new admin account is the most visible result of a successful privilege escalation. Route admin creation and role changes to a monitored log or ticket queue.
- Patch authorization flaws as fast as remote code execution. Once an attacker is an administrator, many products offer a path to the operating system anyway.
For a worked example of these steps on one product, see how to patch Aspera Faspex 5 and audit admin accounts. For the bigger picture of why this product category keeps appearing in exploit reports, read why file transfer servers get exploited.
Testing your own applications for CWE-863
If you build or buy custom applications, ask for role-based access tests. The method is simple: list every action, list every role, and confirm that each role is refused every action it should not perform, through the user interface and directly through the API. Check that IDs in requests are validated against the session owner on every call. Our glossary entry on missing authorization covers the closely related case where the check is absent.