Looking after a Windows server is easy to get wrong when it is one of ten jobs you do. These are the five Windows server security mistakes we see most often in small offices, what each looks like in practice, and how to fix it. Most of the fixes take under an hour.
1. Leaving Remote Desktop open to the internet
This is the most dangerous of the five. A Remote Desktop port forwarded from the router to the server is visible to every scanner on the internet, and password-guessing tools find it quickly. The Security log fills with failed sign-ins (event 4625).
How it happens: someone needed to work from home during the GST or ITR rush, the IT vendor forwarded port 3389 on the router, and it stayed that way. Nobody decided to be insecure; there was a deadline.
Fix: close it, and give remote staff another way in. Prixl Guard Secure Access puts each invited person on an encrypted tunnel to the server so Remote Desktop no longer needs to face the internet, and it works without router changes or a static IP. A VPN or RD Gateway with multi-factor sign-in does the same job. If you cannot do either yet, restrict RDP to known IP addresses on the firewall and the router. Changing the port number only cuts log noise; it is not the fix.
To check whether you are exposed, test the port RDP really uses (it may have been moved off 3389), and test from outside the office, for example from a laptop on a phone hotspot. Testing your public IP from inside the office gives misleading answers on many routers. On the server, this shows the real port:
(Get-ItemProperty -Path 'HKLM:\SYSTEM\CurrentControlSet\Control\Terminal Server\WinStations\RDP-Tcp' -Name PortNumber).PortNumber
Then, from outside, with your own public IP and that port:
Test-NetConnection 203.0.113.10 -Port 3389
If TcpTestSucceeded is True, every bot on the internet can reach it too. Warden alerts you to both fast floods and slow, low-and-slow guessing campaigns, so you know whether it is happening while you fix it.
2. Weak, short or shared passwords
“Admin123”, “Firm@2026”, the firm’s name plus the year: guessing lists hold thousands of these patterns and try them in roughly the order you would expect. If the built-in Administrator account has one of them, compromise is a matter of time. It gets worse when the same password is used on the server, the router and the income-tax or GST portal login, because one leak then opens everything.
Fix:
- Set a minimum password length through policy so it is not optional. We recommend 15 characters, in line with NIST SP 800-63B-4 (2025); a passphrase of a few unrelated words is fine. Windows Server 2016 and 2019 cap the setting at 14, so set 14 there and make 15 the written rule.
- Rename or disable the built-in Administrator, the first username every attack tries, and give each admin a named account.
- Turn on account lockout. Microsoft’s newer default is 10 attempts in 10 minutes, locked for 10 minutes (KB5020282). Note the catch that ties this to the previous point: on Server 2022 and older, the built-in Administrator is exempt from lockout unless “Allow Administrator account lockout” is enabled. If you keep that account, enable it. And on a server still open to the internet, lockout lets attackers lock your staff out, which is why mistake 1 comes first.
3. Not watching sign-in activity
Windows records every sign-in and every failure: successful sign-ins are event 4624, failures are 4625, lockouts are 4740. The evidence of most attacks sits in the Security log as they happen. If nobody reads it, it protects nothing, and the entries scroll away unread. Our guide on how to tell if someone is trying to break in shows what to look for.
Fix: make sure the events are recorded (our audit policy baseline takes ten minutes), then automate the reading. You do not need a SIEM; password-guessing can be spotted from the event log directly. Warden reads these events on the server and emails you when something needs attention, with a plain-English explanation of what happened.
4. Ignoring sign-ins outside office hours
Someone using stolen credentials signs in with a valid username and password, and Windows records a perfectly normal sign-in. What they often cannot hide is when. A Remote Desktop sign-in at 2 AM on a Sunday deserves a look; without anyone watching, it goes unnoticed until Monday, if ever.
Fix: define your working hours and check successful sign-ins outside them. After-hours sign-in monitoring is one of the highest-signal checks you can run. In Warden you set the hours and working days in the dashboard (Settings > Working Hours, for example 9:30 AM to 7:30 PM, Monday to Saturday). Until you save your own hours, Warden records odd-hour sign-ins but does not email them, so do this on day one. Widen the hours during the tax-filing rush if staff work late, rather than learning to ignore the alerts.
5. Not tracking account changes
Once inside, attackers make sure they can come back: they create an account, add it to Administrators, or quietly re-enable a disabled one. Windows logs each move: account created is event 4720, account enabled is 4722, added to a local group such as Administrators is 4732. If nobody sees these, the attacker keeps a back door, and the eventual clean-up means checking every account they might have touched rather than one.
Fix: review net localgroup Administrators now, and get alerted on account changes from here on. Warden emails you about new accounts, admin-group changes, account changes and lockouts, so a new admin account you did not create does not wait for the next audit to be found.
The pattern behind all five
None of these is a missing product, and none needs a security budget. Mistakes 1 and 2 are prevention: closing doors that should never have been open. Our server security checklist and 5 simple steps to secure your Windows server cover them in more depth. Mistakes 3, 4 and 5 are visibility: nobody watching the doors that must stay open.
The prevention side is an afternoon’s work. For the visibility side, start your 30-day free trial of Prixl Warden and let it watch mistakes 3, 4 and 5 for you.

