Blog
data security in applications (sixth pass: the account of someone who already left)

Ex-employee accounts: who can still log in?

Marlon TrettinPublished on 7 min read
Corporate lobby at night with a wall of small lockers glowing teal, a few of them open and spilling amber light onto the polished floor, while a security guard seen from behind checks a glowing panel at the front desk

When someone leaves a company, the departure is recorded in HR, but access stays in every system, each with its own user list. If no account list is tied to the people records, shutting access off depends on someone remembering every one of those systems. That is an architecture problem, not a discipline problem.

Why does a departed person's account stay active?

Because a departure is an event and access is a state spread across systems. HR records the exit in one place. The accounts sit in as many places as the company has systems, each with its own administrator and its own user list.

This is an architectural observation, not a research result. When the departure notice reaches whoever administers the ERP by email, the account is closed in the ERP. A tool one department bought, a supplier's portal, and a shared login are often on nobody's list. The problem is not one person's oversight. It is that no list says which accounts exist for whom.

What happened at the US state government organization?

An attacker got into the network through the account of a former employee that still existed. Advisory AA24-046A, published on February 15, 2024 by CISA, the US cybersecurity agency, and MS-ISAC, describes the case. The account had administrator privileges and, according to the affected organization, was not disabled right after the person left. The credentials likely came from an earlier data breach elsewhere.

With the account, the attacker connected through the VPN and reached two virtualized servers: one running SharePoint and one holding the former employee's own workstation. The SharePoint server stored the credentials of a second account, a global domain administrator, which the attacker used in the on-premises directory and in Azure AD. Neither administrative account had multifactor authentication (MFA).

Once notified, the organization disabled the former employee's account immediately, took both servers offline, and changed the second account's password. CISA's recommendations include a policy and procedure for promptly removing accounts that are no longer needed, especially privileged ones, and a continuous user-management process that makes sure people who have left can no longer reach the network.

The case stacks several layers: a leftover account, high privilege, no MFA, and an administrator password sitting on an ordinary server. Shutting off the account at departure was the only one that depended on the exit notice reaching whoever administers the account. That reading is ours: CISA does not rank the failures by weight.

What does NIST ask for at termination and role change?

NIST asks for a defined time period to disable access, and a link between account management and the departure of people. The SP 800-53 catalog, revision 5, is a US technical reference. For a midsize private company, it is usually a benchmark, not an obligation.

Control PS-4, on personnel termination, says the organization must disable system access within a period it defines and revoke the person's credentials. Control AC-2, on account management, says it must notify account managers, within a defined period, when someone is terminated or transferred, align account management with the termination and transfer processes, and review accounts at a defined frequency. The enhancement AC-2(3) asks for disabling accounts that are no longer associated with a person or that have been inactive for a defined period.

Two details matter. NIST does not fix the time period; each organization picks its own, so the useful question is what ours is and who checks it. And the text covers transfers, not only exits: someone who moves to another department takes the old department's access along unless someone removes it. The catalog also provides for automated mechanisms to notify and disable access at termination (PS-4(2)), but only at its higher protection levels.

How does this show up in a small or midsize operation?

As a list that nobody has. The example below is a hypothesis, not a measurement.

Picture a distributor with 60 people and seven systems with logins: ERP, CRM, email, a shared drive, the bank portal, the freight carrier portal, and time tracking. When a salesperson leaves, HR emails IT. IT closes the email account and the ERP, the two it knows. The CRM is administered by Sales, who were not told. The carrier portal has one login used by three people, and its password never changes.

If eight people leave each year and one account slips through at each exit, three years later there are 24 active accounts belonging to people who no longer work there. The arithmetic is ours and illustrative, but the mechanism is the one in the US case: the account does not need to be forced, it only needs to exist. And every forgotten account remains an open door.

In the Repap On case, document visibility is defined by role and department, which prevents unauthorized access to confidential documents. The logic helps here: when access depends on role and department, moving a person to a new function is one change in one place, not seven manual adjustments.

What can you do so offboarding does not depend on memory?

This part is our reading, not a conclusion from the sources. Five decisions, in this order.

First, one list of accounts per person. For each system: who has access, at what level, and who owns the system. It fits in an afternoon, and the first version will have gaps. It already shows where they are.

Second, the HR event triggers the shutdown. A departure or a role change creates a task for each system owner, with a deadline and a confirmation. The deadline is the company's decision, as in NIST, and whatever is unconfirmed stays visible instead of vanishing into an inbox.

Third, centralized sign-in where possible. The more systems accept the same corporate login, the fewer accounts there are to disable. What cannot join goes on an exceptions list, each with an owner.

Fourth, specific handling for privileged and shared accounts. Multifactor authentication on remote access and administrator accounts, individual names instead of collective logins, and a password change when someone who knew it leaves. Administrator credentials are not stored on an ordinary server.

Fifth, a periodic review. Compare the account list with the active payroll. An account with no matching person is a finding, not a detail, and how often to check is the company's call. For the access levels themselves, see access control: who sees what in your systems. For knowledge that only one person holds, see does your internal system depend on one person?.

Frequently asked questions

Will integrating with the systems we already use be complicated?

Not every system needs automatic integration at the start. The most critical ones, such as finance, email, and the ERP, go first, and the rest become tasks the owner confirms. Integration comes later, where the volume of departures justifies it.

My team tracks access in a spreadsheet. Is that enough?

As a starting point, yes. The limit is manual upkeep: the spreadsheet is right only if someone corrects it at every exit and every role change. When it stops matching the active payroll, that is the sign the list needs to be tied to the people records.

I have no IT team for this. Who does it?

The first step does not need IT: HR and each system owner list the accounts and agree on a deadline. The register tied to HR is work an outside partner can design once, and after that the routine belongs to the company.

The next step

If you cannot say how many accounts of people who have left are still active today, that is one of the questions the Operational Architecture Diagnostic, a 30-minute conversation, helps open.

Related case studies

Marlon Trettin

Marlon Trettin

Founder of Yowpi · 25+ years of software engineering

LinkedIn
Next step

Did you recognize your operation in this article?

Book the Operational Architecture Diagnostic: 30 minutes to map where your operation's bottleneck is. No proposal in the first conversation.

Book a diagnostic