Access in WBAMS is decided in two places: who has an administrator account, and which role that account carries. Everything else — two-factor, sign-in protection, fraud checks — hardens those two decisions.
Administrator accounts
Each administrator has their own account. Shared logins destroy the audit trail, because every entry in the administrator log then names the same person. Invite a new administrator by email; the invitation is accepted on a page of its own where they set their own password.
A single super-administrator account whose password is lost is an outage. Verify the second account can actually sign in rather than assuming it can.
Roles and the permission list
A role is a named set of the 177 permissions, presented under the administration area’s own headings so a role can be reasoned about area by area rather than as one undivided list. Each heading has Check All and Uncheck All.
- Start from the narrowest role that could do the job, not from a copy of the full administrator role.
- Grant whole areas where the job needs them; grant individual permissions where it does not.
- Test the role by signing in as a non-super-admin account holding it. Reading the checkboxes is not a test.
- Review roles after staffing changes rather than on a schedule.
Pages do not appear greyed out; they are absent from the menu and return access denied if addressed directly. An administrator reporting a missing page is usually reporting a permission, not a bug.
Who may change the website
Since 1.12.0 the Content Manager has five permissions of its own, under Help Center and content, so a role can be given one part of the website. Manage Content still allows all of it, so no role loses anything at the upgrade, and a role that held every permission is given the five new ones too.
The Content Manager's overview and traffic screens open with any of the five, and its own menu shows only the parts a role has. A screen that belongs to none of them still needs Manage Content. A refused screen or save names the permission it needs.
Two-factor authentication
Two-factor applies to administrator sign-in. Enable it, then enroll accounts individually. Keep recovery arrangements for an administrator who loses their device, and confirm the recovery path works before you need it.
Security questions
Security questions are used for account recovery. Maintain the question list deliberately: a question whose answer is discoverable from a public profile is not a control.
Sign-in protection
Repeated invalid sign-in attempts ban the originating address for a period. Bans are visible and removable under blocked addresses.
A monitor, uptime check or crawler that hits the administrator sign-in URL will ban itself, and can ban an office address that shares the same egress IP. Point health checks at a page that does not authenticate.
Client sign-ins and password resets
Since 1.11.0 wrong client sign-ins are limited wherever they come from — the sign-in page, the cart's sign-in and the client API. After 5 wrong attempts for one account name, or 20 from one address, within 15 minutes, even the right password is refused with the usual "incorrect" message until the 15 minutes have passed. A new password set through the reset email lifts the limit at once. Password reset emails are limited too: 3 per address asked for and 10 per visitor in 15 minutes, after which the page answers as usual and sends nothing. Nobody is banned for longer, and nothing on screen tells a limited attempt from a wrong one.
An administrator's VerifySignIn API call is not limited. Details are in Operations and security.
Sessions
Signing in, as an administrator or as a client, replaces the session id and ends the old one. The session cookie is SameSite=Lax and PHP's strict session mode is on, so a session id planted in a browser before sign-in is never adopted. See Sessions.
Blocked addresses and emails
Blocked IP addresses and blocked email addresses are separate lists. The IP list governs who may reach the sign-in and ordering flows; the email list governs which addresses may register or open tickets. Both are reviewed rather than accumulated — an entry added during an incident and never removed becomes a support mystery months later.
Fraud checks
Fraud checking runs against orders at the point they are placed. Configure the provider, decide what a failed check should do to the order, and make sure somebody is reviewing what it flags. A fraud rule nobody acts on is a queue, not a control.
Sign-in integrations
OpenID and third-party sign-in let customers authenticate with an existing identity provider. Configure the provider credentials, decide whether an account may be created on first sign-in, and confirm the callback address matches exactly — a trailing slash difference is the usual cause of a failed round trip.