The ticket came in on a Thursday afternoon, buried about a third of the way down my queue. Subject line: "Cannot log in to my account." Priority: normal. The user had submitted it two days prior and said they had been locked out since Monday morning. They had not had access in almost four days.
I pulled up the account in our identity platform and found it immediately: an active account, no lockout flag, password just needed a reset. Standard. I had done this exact sequence probably a dozen times that week. I should have been in and out in four minutes.
Instead, I paused on a detail I almost scrolled past.
The timestamp that did not add up
The account showed a last successful sign-in on Monday at 8:52 AM. The user had said they lost access on Monday morning. That overlap seemed fine at first. But then I looked more carefully at the full sign-in log rather than just the summary field. That is the habit I want to talk about: the two-second decision to open the full log instead of just processing the reset.
What the full log showed was not what I expected. There were successful authentications on Monday evening, all day Tuesday, and all of Wednesday. After the user told me they had not been able to get in. And every one of those logins came from an IP address in a city neither of us had any connection to, mapping to a location in Eastern Europe. The user's normal sign-in pattern was entirely different: same general region, same device fingerprint, consistent timing.
I did not touch the password reset yet.
Before I changed anything, I picked up the phone
The user confirmed what I had already suspected. They had not traveled. They had not shared credentials with anyone. They had not logged in successfully from anywhere since Sunday evening. When I read out the city and the timestamps, there was a pause that told me everything I needed to know.
What had happened, as best we could reconstruct it, was this: the user had clicked a link in what appeared to be an internal notification email about two weeks earlier. The page it led to was a convincing replica of our SSO login screen. They had entered their credentials. Within hours, someone had the password and started using the account quietly. The attacker changed the password at some point on Monday morning, which is what locked the user out and generated the support ticket. By the time I was looking at the account on Thursday, the unauthorized access had been running for three days straight.
This is no longer a password reset
I revoked all active sessions on the account immediately, which ended the unauthorized access in that moment. Then I escalated to the security team and opened an incident, with the timestamps, the IP data, and the user's confirmation that none of those logins were theirs.
What followed was an audit of everything that account could see. The user had standard employee permissions: read access to two shared drives, their own email inbox, and a handful of internal SaaS tools. None of that sounds alarming until you look at what was actually in the inbox during those three days.
There was an active email thread about an upcoming wire transfer to a vendor. Full details: amounts, account numbers, the approval chain, the scheduled date. The attacker had not replied to any emails, had not moved any files, had not done anything that would have triggered an alert. They were just reading. Watching. Waiting, possibly, for the right moment.
Finance was notified the same afternoon. The wire transfer was held pending a re-verification of the destination account through a separate channel. When we called the vendor directly to confirm the banking details, the information matched what was in the thread, so in this case the transfer was legitimate. But the week of delay and the verification process cost real time on both sides, and the possibility that someone had been sitting in that inbox while that conversation was happening was not a comfortable one.
The question the incident kept raising
The technical piece of this was not especially complicated. Revoke sessions, force reset, audit access, verify the wire, brief the security team, document everything. All of that has a clear process once you know you have an incident.
What stayed with me was a simpler question: how many password resets had I processed that week without opening the sign-in logs? A dozen, conservatively. Probably more. Almost all of them were exactly what they looked like: a user who forgot their password or hit a lockout. But "almost all" is not the same as "all," and the full log takes maybe fifteen seconds to scan once you know what you are looking for.
What I was looking for, and now always look for, is whether the sign-in history is consistent with the user's story. If someone says they have not been able to log in since Monday and the account shows successful logins Tuesday and Wednesday, those two things cannot both be true. Either the user is mistaken about when they lost access, or someone else has been in the account. Both of those are worth a question before you close the ticket.
What changed after this
I added a step to my own password reset process. Before touching a password on any account that has been active in the last 48 hours, I open the full sign-in log and check whether the timestamps and locations are consistent with what the user reported. It adds maybe thirty seconds to the ticket. In hundreds of resets since then, it has flagged one other account with a login pattern that did not match, which turned out to be a different user sharing credentials with a contractor (a problem in its own right, but not a breach).
I also documented the check as a recommended step in our internal runbook for password resets. Not every team member will do it on every ticket. But having it written down means that when someone is onboarding or working a high volume of resets under time pressure, the question is at least in front of them: does the sign-in history match what the user is telling you?
The attacker in this case got three days of quiet access to an employee inbox because a password reset was sitting in a queue. The fix was not new software or a new policy. It was one extra screen before the close.
A support ticket is a window into the account as it actually is, not just as it was reported. The two seconds it takes to verify that view is part of the job.
I think about this case whenever I am working a queue under pressure and feel the pull to process and move on. The tickets that look routine usually are. But the sign-in log does not care whether the queue is long. It just tells you what happened.
Open it before you close the ticket.
Questions about access security or support operations? Get in touch.