About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
A raw count of failed sign-ins is not a finding. Every Windows server records some failures, because people mistype passwords and software retries with stale ones. What matters is the shape: many failures against one account, failures spread across many account names, failures from a source that should never be signing in, or a run of failures followed by a success. This guide explains how to read those shapes, and which items in a report to chase first.
It is for business owners and IT administrators who receive monthly security or service reports for Windows servers. It relies on Microsoft's documentation of three event IDs. Event 4625 records a failed sign-in. Event 4740 records an account lockout. Event 4624 records a successful sign-in. Nothing in this guide is specific to any one provider's report format.
Why it matters
Sign-in attempts are the front door. Microsoft's own documentation on lockout settings explains the threat plainly: "Brute force password attacks can be automated to try thousands or even millions of password combinations for any or all user accounts." A report that only shows totals hides whether that is happening.
The opposite risk is wasted effort. The same pages describe failures that are "typically not a security issue" but an infrastructure or availability problem. Telling the two apart saves your team and your provider from chasing noise while a real pattern goes unread.
What a failed sign-in event records
Microsoft's page for event 4625 says the event "is logged for any logon failure" and that it generates "on the computer where logon attempt was made". The fields that matter for reading a report are:
- Account name tried, under "Account For Which Logon Failed". This is the name typed, which may not exist.
- Logon type, a number saying how the sign-in was attempted. The page lists them: type 2 is interactive at the machine, type 3 is network, type 5 is a service started by the Service Control Manager, and type 10 is "RemoteInteractive": "A user logged on to this computer remotely using Terminal Services or Remote Desktop."
- Source network address and workstation name, which say where the attempt came from.
- Status and sub status, coded reasons the sign-in failed.
- Authentication package, such as NTLM or Kerberos.
The same event on the Microsoft page can also show the process that made the attempt. Our reading: a failure count with no breakdown by these fields cannot be judged. Ask your provider whether the report is built from these fields or only counts events.
Reasons that matter, and reasons that are noise
The status code is the most useful field. Microsoft's monitoring recommendations for event 4625 list a number of status values. Our grouping of them is below.
Possible attack signals
- Unknown account names (status 0xC0000064, "User logon with misspelled or bad user account"). Microsoft says that, "Especially if you get several of these events in a row, it can be a sign of a user enumeration attack." That means someone is guessing which usernames exist.
- Wrong password, repeated (0xC000006A). Microsoft advises to "Especially watch for a number of such events in a row", for critical accounts or service accounts.
- Outside permitted use: sign-in outside authorised hours (0xC000006F), from an unauthorised workstation (0xC0000070), to a disabled account (0xC0000072), to an expired account (0xC0000193), or where the account lacks the logon right (0xC000015B). Each can be an honest mistake or a probe of an old account. Our reading: any of these on an account that should not be in use is worth a question.
Probably routine or infrastructure
- "There are currently no logon servers available to service the logon request" (0xC000005E), and "the Netlogon service was not started" (0xC0000192). Microsoft says each "is typically not a security issue" but may be an infrastructure or availability issue.
- Single wrong-password failures spread across ordinary users, one or two per person, are often mistyping (our reading). Check the pattern before you call them routine: password spraying tries a few common passwords against many accounts, so each account stays under the lockout threshold. See "Low and slow" below.
Account lockout status (0xC0000234) appears when a sign-in fails because the account is already locked. Our reading: counts of these can be inflated by one stuck process retrying.
Lockouts: what 4740 adds
Event 4740 "generates every time a user account is locked out". Its useful fields are the account locked out and the caller computer name, the machine the attempts came from. Microsoft recommends monitoring every lockout for high-value accounts, and alerting when the caller computer is unexpected.
How often lockouts occur depends on settings. Microsoft's pages say the account lockout threshold is the number of failed attempts that locks an account, from 1 to 999, with 0 meaning never lock. The Windows security baselines it cites recommend 10 as an acceptable starting point. The pages also set out a trade-off that changes how to read a report:
- A threshold above zero stops guessing but allows lockout-by-attack. Microsoft warns that "an attacker can easily lock any accounts for which the account name is known". A burst of lockouts may therefore be a nuisance attack rather than a break-in.
- A threshold of zero never locks. Microsoft says this "might go undetected if a robust audit mechanism isn't in place". In that setup, the failure counts are the only warning, so they matter more.
- Retries can inflate counts. Microsoft notes that Kerberos "can automatically retry" sign-in attempts that count toward the threshold, and that when "a connection drops repeatedly" in an application, all subsequent failed sign-ins count toward the threshold.
- The built-in Administrator account depends on the Windows update level and policy. An older Microsoft threshold page says the account "is excluded from this policy", but a later Microsoft support article, KB5020282, says that from the 11 October 2022 or later cumulative updates a local policy, "Allow Administrator account lockout", is available to lock out the built-in local Administrator account. On existing computers it must be set to Enabled by local or domain policy. On new computers on Windows 11 version 22H2, or computers that included those updates before first setup, it is set by default; a computer set up first and updated later does not get it by default. The article says the new lockout "only affects network logons, such as RDP attempts", and that console logons are still allowed during the lockout. So whether the built-in Administrator generates lockouts and 4740 events on a given server depends on its patch level, how it was set up and what policy applies. Ask the provider for the effective setting on each server, and read both the 4625 failures and any 4740 lockouts for that account.
Two companion settings control timing: lockout duration (Microsoft advises approximately 15 minutes) and the time after which the failure counter resets (its baselines recommend 15).
When a zero means nothing
A report that shows zero failures does not prove nobody tried. The Microsoft pages tie each event to an audit subcategory: event 4625 to "Audit Account Lockout and Audit Logon", 4740 to "Audit User Account Management" and 4624 to "Audit Logon". Our reading: if auditing for failures is not switched on for the server, the log has nothing to count and the report shows a clean zero. Ask whether auditing is enabled for these subcategories, on which servers, and how long the log is kept before it overwrites.
Another gap is scope. The 4625 page says the event generates on domain controllers, member servers and workstations, on the machine where the attempt was made. A report covering only some servers gives only part of the picture.
What to chase first
Our recommended order, applied to a monthly report:
- Failures against privileged and service accounts. Microsoft recommends monitoring all 4625 events for local accounts and service accounts, because "these accounts should not be locked out". Chase any pattern here first.
- Failures followed by a success. A run of failures against one account, then a successful 4624 event for that account from the same source address, is the pattern that most needs a check. The 4624 event records successful sign-ins, with the logon type and source address (our reading: matching by account and source is how to link the two).
- Many different account names from one source. That fits the user-enumeration warning.
- Failures from unexpected places. Compare the source addresses against the networks and machines your staff actually use. If the server accepts Remote Desktop sign-ins, type 10 failures from unknown addresses are the ones to ask about (type 10 is the Remote Desktop type in Microsoft's table).
- Account names that are not on your user list. Compare every account name in the failures against your current staff and service accounts. Names of former staff or default names point to old accounts or guessing.
- Low and slow patterns. Microsoft's password-spray guidance describes "low and slow" indicators as the case where "thresholds for account lockout or bad passwords aren't being hit". It lists failures that follow the order of the Global Address List, failures with repeated attributes such as the same address block, and timing, because "automated sprays tend to have a more regular time interval between attempts". The guidance is written for AD FS and Microsoft Entra ID, so treat it as a pattern to look for rather than a Windows Server event recipe (our reading). Before you class a spread of one-or-two failures per user as mistyping, check for a shared source address, regular timing, many accounts in the same short period, and any success after the failures.
- Noise last. Single failures per user that pass that check, and infrastructure statuses.
Read-only checks
These steps read the log and change nothing. Security event contents include account names and addresses, so handle the output as sensitive. The following PowerShell command was not tested by the author. It lists up to 50 failed sign-in events from the Security log for the last day. According to Microsoft's Get-WinEvent reference, the cmdlet supports a hash-table filter with LogName, ID and StartTime keys and a MaxEvents limit, and PowerShell may need to run as Administrator.
PowerShell:
Get-WinEvent -FilterHashtable @{ LogName='Security'; Id=4625; StartTime=(Get-Date).AddDays(-1) } -MaxEvents 50
Changing Id=4625 to Id=4740 lists lockouts. An empty result can mean either that nothing occurred or that auditing is not enabled (our reading).
Questions to put to your provider
- Does the report count events or break them down by account, source address, logon type and status?
- Is failure auditing enabled for Audit Logon and Audit Account Lockout on every server in scope?
- What is the lockout threshold, is "Allow Administrator account lockout" effectively enabled for the built-in Administrator on each server, and does anything differ for service accounts?
- What counts as an alert, and who is told when a lockout hits a privileged account?
- How long are the logs kept before they are overwritten?
- Can we compare the account names in the failures with our user list each month?
A worked example (illustrative)
The following is invented. A monthly report says 3,200 failed sign-ins. The firm asks for a breakdown. About 300 failures come from 40 real staff accounts. A further 2,800 come from one external address, trying 600 different names, nearly all with the "bad user account" status. The last 100 are against a service account from a machine the firm believed was retired, using an old password.
The firm treats the 2,800 as an enumeration attempt and asks the provider to check that no attempt succeeded and that Remote Desktop is not open to the whole internet. For the 300 staff failures, it checks before calling them mistyping: the source addresses are the firm's office and VPN ranges, the failures are spread through working hours, and no account has a success from an unexpected address. Had they shared one outside address or arrived at regular intervals, they would have gone on the chase list as a possible spray. For the service account, it identifies the host still sending the old password, which turns out to be a retired-but-still-running scheduled job on a forgotten test machine. It then stops that job and updates or removes the stored password.
Read next
Related articles you may want next.
- TroubleshootingAccount Keeps Locking Out: Find the Source From Event 4740 to the ProcessAccount keeps locking out? Read event 4740 on the PDC emulator, follow 4771, 4776 and 4625 to the source computer, then find the stale credential there.Read article
- TroubleshootingBefore Restricting NTLM: Inventory Who Uses It and Why Kerberos Was Not UsedAudit NTLM before blocking it: enable the Restrict NTLM audit policies, read the NTLM Operational log and event 4624, fix IP and SPN fallbacks, then restrict in rings.Read article
- TroubleshootingSMB Signing Now Required: NAS and Old Client Access FailuresNAS share fails after Windows 11 24H2 or Server 2025? Tell signing (STATUS_INVALID_SIGNATURE) from blocked guest access, audit with SMBClient events, fix the device first.Read article
Risks and limits
- Event content and status codes differ by Windows version; Microsoft's pages are for Windows 10 and Windows Server and mostly carry older "last updated" dates (the 4625 page, updated 29 March 2026, is the exception).
- Our grouping into attack signals and routine items is a reading of Microsoft's monitoring notes, not a published classification.
- A report from a different tool may count different things.
- The logs can show failures only if auditing was on at the time.
- The article does not cover other sign-in systems such as cloud identity services.
Next step
Compare the account names appearing in your failed sign-in counts against your current user list, and ask which servers have failure auditing switched on. If you are reviewing a hosted service, the Managed Services page is a starting point for that conversation; ask for the scope in writing.
Sources
- 4625(F): An account failed to log on, Microsoft Learn, last updated 29 March 2026.
- 4740(S): A user account was locked out, Microsoft Learn, last updated 7 September 2021.
- 4624(S): An account was successfully logged on, Microsoft Learn, last updated 7 September 2021.
- Account lockout threshold, Microsoft Learn, last updated 2 November 2018.
- Account lockout duration, Microsoft Learn, last updated 16 August 2021.
- Reset account lockout counter after, Microsoft Learn, last updated 2 November 2018.
- KB5020282: Account lockout available for built-in local administrators, Microsoft Support, no date displayed (October 2022 cumulative updates), read 16 November 2026.
- Password spray investigation, Microsoft Learn, last updated 4 November 2026 as displayed, read 16 November 2026.
- Get-WinEvent, Microsoft Learn, date unavailable.
All accessed 15 November 2026, except the two sources dated 16 November 2026 above.
Learning notes
- A failed sign-in count means little without account, source, type and reason.
- Repeated "unknown user" failures suggest someone guessing names; repeated "wrong password" on important accounts deserves a look.
- Lockouts can be caused by an attacker or by one stuck application.
- Whether the built-in Administrator account locks out depends on the update level and the "Allow Administrator account lockout" policy, so ask for the effective setting and read its failures as well as its lockouts.
- A few failures per user can still be a password spray; check source, timing and any success before calling it noise.
- A zero can mean nothing happened, or that auditing is off.
This article was originally published on .