Skip to content

Event ID 4625: Failed Logon Causes, Codes, Detection

Event ID 4625: Failed Logon Causes, Codes, Detection

A working reference for the Windows failed logon event: what each field means, how to read the Status codes, and how to tell a typo from a spray.

Event ID 4625 is the Windows Security log entry written whenever a logon attempt fails. It names the account that was tried, the logon type, the source address and a Status and Sub Status code that explains the failure. A single 4625 usually means a typo. Many of them across different accounts from one source address is the pattern to alert on.

This cheat sheet covers the status codes, the logon types, a Get-WinEvent pull, a KQL query and a Sigma rule. It sits next to our guide to Sysmon event IDs, which covers process and network telemetry, and the Sigma and YARA rules page where we keep detection content.

If failed logons are climbing on a domain controller during a live case, start with the incident response guide or the first 60 minutes of Active Directory ransomware containment. When a run of failures ends in a success, endpoint work such as hunting living-off-the-land activity in EDR telemetry takes over.

Key takeaways

  • Windows logs Event ID 4625 on the computer where the logon attempt was made, which includes domain controllers, member servers and workstations [1].
  • Read Status and Sub Status together. Microsoft singles out 0xC0000064 (bad username), 0xC000006A (bad password) and 0xC000006D (bad username or authentication information) [1].
  • Logon type shows the path in: type 2 is interactive, type 3 is network, type 10 is Remote Desktop or Terminal Services [1].
  • A password spray tries one password or a short list against many accounts, so count distinct accounts per source address instead of failures per account [3].
  • MITRE notes that LDAP and Kerberos attempts are less likely to trigger a 4625, so an empty 4625 log does not prove a quiet network [3].
  • The SigmaHQ rule for the spray pattern carries the status "unsupported", so test any threshold in your own environment before alerting on it [4].

Event ID 4625 is the Windows Security event that records a failed logon, and Microsoft documents it across the Windows estate, from domain controllers to workstations [1]. Windows writes it on the computer where the attempt was made, so a failed local logon on a workstation lands in that workstation's own log [1].

Each record carries the account that was tried, the logon type, the workstation name, the source address, the calling process and a Status and Sub Status pair that gives the failure reason [1]. Together those fields answer who, how, why and from where.

Microsoft's monitoring guidance also recommends watching every 4625 for local accounts and service accounts, since neither should lock out in normal use [1]. MITRE adds a caution from the attacker side: sprays over LDAP or Kerberos are less likely to produce this event than failed SMB connections [3].

What does Event ID 4625 mean?

Event ID 4625 is a Windows Security log event described as "An account failed to log on". Windows generates it on the computer where the logon attempt happened, so it appears on workstations, member servers and domain controllers. Every record names the target account, logon type, source address, process and the failure codes that explain why the attempt was refused.

Microsoft files the 4625 event under two audit subcategories, Audit Logon and Audit Account Lockout [1]. If the event never appears, check those audit settings before you suspect the attacker is quiet.

The event groups its fields into blocks. It reads faster once you know what each block answers:

  • Subject: the account that reported the failure.
  • Account For Which Logon Failed: the account named in the attempt.
  • Failure Information: Failure Reason, Status and Sub Status.
  • Process Information: the process ID and name that made the attempt.
  • Network Information: workstation name, source address and source port.
  • Detailed Authentication Information: logon process, authentication package (NTLM, Kerberos or Negotiate) and key length.

Those blocks come straight from Microsoft's field descriptions [1]. The diagram below turns them into a five-step reading order for a 4625.

Five checks for reading Event ID 4625 A vertical flow of five numbered boxes: logon type, status and sub status, target account, source address and workstation, and process name. Arrows run from each box to the next. 1. Logon Type How did the attempt arrive? 2. Status and Sub Status Why did it fail? 3. Account For Which Logon Failed Who was targeted? 4. Source address and workstation Where did it come from? 5. Process name Which program asked?
Figure 1. A reading order for Event ID 4625, built from Microsoft's field descriptions. Conceptual diagram, not a capture of a real log.

One more detail from Microsoft's page: its sample 4625 is an account lockout, with Status 0xC0000234 and the reason "Account locked out" [1]. A 4625 does not always mean a wrong password.

What do the Status and Sub Status codes in Event ID 4625 mean?

The Status and Sub Status fields in Event ID 4625 hold NTSTATUS values that give the reason a logon failed. 0xC0000064 points to a bad username, 0xC000006A to a bad password, and 0xC000006D to a bad username or bad authentication information. Read both fields, because Microsoft's guidance applies each code to Status or Sub Status.

Microsoft sorts these codes into security signals and operations problems. It describes 0xC000005E and 0xC0000192 as typically not a security issue but an infrastructure or availability issue [1].

The table lists the codes that Microsoft's monitoring guidance calls out for this event, with a short reading for each. The right column is analysis, not Microsoft's wording.

Code Microsoft's description How to read it
0xC0000064 Logon with a misspelled or bad user account Microsoft says several in a row can indicate user enumeration
0xC000006A Logon with a misspelled or bad password Real account, wrong secret. Watch runs against critical and service accounts
0xC000006D Bad username or authentication information Broad wording. Check Sub Status for the specific cause
0xC000006F Logon outside authorized hours A logon-hours restriction did its job
0xC0000070 Logon from an unauthorized workstation Right account, wrong machine
0xC0000072 Logon to an account disabled by an administrator A disabled account is still in use
0xC0000193 Logon with an expired account An account past its expiry date
0xC0000234 Account locked out (the typical value on Microsoft's page) Look for the failures that came before the lockout
0xC000015B User not granted the requested logon type at this machine A logon right is missing, often a configuration issue
0xC0000413 Machine is behind an authentication firewall and the account is not allowed An authentication firewall policy blocked it
0xC000005E No logon servers available Send to operations first
0xC0000192 Netlogon service not started Send to operations first

Microsoft's descriptions come from its monitoring table for the event, which applies each code to Status or Sub Status [1]. The page also points to a separate NTSTATUS reference for codes outside this list.

A single 4625 is a typo until the source address, the account list or the timing says otherwise.

What do logon types 2, 3 and 10 tell you about a failed logon?

The Logon Type field in Event ID 4625 shows how a failed attempt reached the machine. Type 2 is a local interactive logon, type 3 is a network logon from another computer, and type 10 is a Remote Desktop or Terminal Services logon. Types 3 and 10 deserve the closest look, because both come from other machines.

Microsoft documents nine logon types for this event [1]. MITRE lists SMB and RDP among the services attackers commonly spray [3], which is why network and remote interactive failures draw the most attention.

Type Microsoft's description Reading a failure
2 Interactive A user logged on to this computer Someone at the console typed a wrong secret
3 Network A user or computer logged on from the network The main type for remote guessing. Check the source address
4 Batch Processes running for a user without direct intervention A scheduled job with an old password or a removed right
5 Service A service was started by the Service Control Manager A service account whose password or rights changed
7 Unlock The workstation was unlocked A wrong password at the lock screen
8 NetworkCleartext Network logon where the password reached the authentication package unhashed Find the application behind it
9 NewCredentials A caller cloned its token and used new credentials for outbound connections The outbound credentials failed, not the local ones
10 RemoteInteractive Logon through Terminal Services or Remote Desktop Remote Desktop guessing, or a mistyped password on a real session
11 CachedInteractive Logon with locally stored network credentials, domain controller not contacted Cached credentials failed without a domain controller check

Mismatches are the useful signal. Microsoft suggests watching for a batch or service logon type used by a member of a domain administrative group [1]. Nobody is typing at a service logon.

Logon type 3 alone does not name the protocol. Read the Authentication Package field, which shows NTLM, Kerberos or Negotiate [1].

How do you pull Event ID 4625 events with Get-WinEvent?

Get-WinEvent with the FilterHashtable parameter pulls Event ID 4625 records from the Security log faster than piping every event into Where-Object. Microsoft lists LogName, Id and StartTime among the accepted keys. Run the command below in an elevated PowerShell session and group the results by source address to see the noisiest senders first.

Microsoft states that FilterHashtable improves performance on large event logs compared with filtering after the pipeline [2]. The hash table keys used here, LogName, Id and StartTime, all appear in Microsoft's key table [2].

Get-WinEvent -FilterHashtable @{
    LogName   = 'Security'
    Id        = 4625
    StartTime = (Get-Date).AddHours(-1)
} | ForEach-Object {
    $d = @{}
    ([xml]$_.ToXml()).Event.EventData.Data |
        ForEach-Object { $d[$_.Name] = $_.'#text' }
    [pscustomobject]@{
        Time      = $_.TimeCreated
        Account   = $d.TargetUserName
        LogonType = $d.LogonType
        Status    = $d.Status
        SubStatus = $d.SubStatus
        Source    = $d.IpAddress
    }
} | Group-Object Source |
    Sort-Object Count -Descending |
    Select-Object -First 10 Count, Name

The field names in that script follow the event XML in Microsoft's 4625 documentation [1]. Print one raw event with ToXml() on your own build first, then adjust if a name differs.

How do you tell a typo from brute force or password spraying in Event ID 4625?

Event ID 4625 patterns separate by how many accounts fail and how fast. A typo is a few failures on one account, then a success. Brute force piles many passwords onto one account. Password spraying tries one password or a short list across many accounts, and attackers can slow it down deliberately to stay under alert thresholds.

Three failed logon patterns in Event ID 4625 Three panels of dots on a timeline. A typo shows two failures on one account followed by one success. Brute force shows ten failures in a row on one account. Password spray shows single failures spread across five accounts over time. Typo One account, a few failures, then success acct A Brute force on one account Many passwords, one target, tight timing acct A Password spray Few passwords, many accounts, spread out acct A acct B acct C acct D acct E failed logon (4625) successful logon Time runs left to right.
Figure 2. How the three patterns look when plotted by account and time. Conceptual diagram, not captured from a real environment.

The patterns map to ATT&CK like this:

  • Typo: not an attack. Expect a successful logon soon after the failures.
  • Brute force: many passwords against one account. MITRE lists the matching sub-technique as Password Guessing, T1110.001 [3].
  • Spray: MITRE catalogs it as Password Spraying, T1110.003, which tries one password or a small list against many accounts to avoid the lockouts a single-account attack would cause [3].

Password spraying inverts the usual guessing pattern. MITRE describes it as one password, or a short list of common passwords, tried against many accounts so that no single account collects enough failures to lock [3]. Attackers can also slow it down on purpose.

MITRE records APT28 tooling that made about four attempts per hour per targeted account over several days or weeks, and a Quad7 variant that made a single sign-in attempt within a 24-hour period [3]. Per-account failure counts never see that pacing. A count of distinct accounts per source address does, which matches the shape of MITRE's detection strategy: many failures using one password, or a small set, across many accounts in a time window [3].

The 4625 log is only one place those failures can appear. MITRE notes that LDAP and Kerberos attempts are less likely to trigger it, so an empty 4625 log does not rule out a spray [3].

A quiet 4625 log is not a quiet network. LDAP and Kerberos sprays are less likely to write one.

What detection query or Sigma rule catches repeated Event ID 4625 failures?

A useful Event ID 4625 detection counts distinct accounts per source address in a short window, not failures per account. The SigmaHQ rule for this idea filters to network logons, drops events with no source address, and fires above ten distinct usernames per address in 24 hours. Treat that threshold as a starting point to tune.

Those conditions come from the SigmaHQ rule "Multiple Users Remotely Failing To Authenticate From Single Source" [4]. The rule sits in the repository's unsupported folder and carries the status "unsupported" [4]. Use it as a design reference, not a drop-in rule.

The KQL below adapts that logic for a Microsoft Sentinel style SecurityEvent table. The one-hour window and the threshold of ten are analysis, not part of the SigmaHQ rule. Check the column names against your own schema before you run it.

SecurityEvent
| where TimeGenerated > ago(1h)
| where EventID == 4625 and LogonType == 3
| where IpAddress !in ("-", "", "::1", "127.0.0.1")
| summarize
    FailedAttempts = count(),
    DistinctAccounts = dcount(TargetUserName),
    SampleAccounts = make_set(TargetUserName, 5)
    by IpAddress
| where DistinctAccounts >= 10
| order by DistinctAccounts desc

The Sigma rule below is the selection half only. It matches network 4625 events that carry a source address, and it leaves the counting to a correlation rule or your SIEM, such as the query above. It has not been tested against a backend.

title: Failed Network Logon With Source Address
id: 3c0bc533-84da-4107-9bd9-554e8744b186
status: experimental
description: Base selection for counting distinct accounts per source address in failed network logons (Event ID 4625).
tags:
    - attack.credential-access
    - attack.t1110.003
logsource:
    product: windows
    service: security
detection:
    selection:
        EventID: 4625
        LogonType: 3
    filter_no_source:
        IpAddress: '-'
    condition: selection and not filter_no_source
falsepositives:
    - Terminal servers, jump servers and Citrix farms
    - Workstations with frequently changing users
level: informational

Count distinct accounts per source address. Sprays hide in the per-account failures nobody counts.

How do you tune out noisy Event ID 4625 sources and handle the real ones?

Tuning Event ID 4625 alerts starts with knowing which sources may fail in bulk. The SigmaHQ rule lists terminal servers, jump servers, Citrix farms and workstations with frequently changing users as false positives. Allowlist those by name with an owner, keep the rule live on everything else, and review the allowlist on a fixed schedule.

That false positive list comes from the SigmaHQ rule itself [4].

Service accounts need the opposite treatment. Microsoft recommends monitoring every 4625 for service accounts, because they should not lock out or fail to function [1]. Fix the cause instead of muting the alert.

Noise source What it looks like in 4625 Tuning move
Jump and terminal servers Many distinct accounts from one address Allowlist by name, with a named owner
Service accounts One account, steady repeats, often type 4 or 5 Do not suppress. Fix the expired password or missing right
Stale saved credentials (old mapped drive, saved task) Same account and workstation, regular rhythm Find the program in the process fields, then update the stored credential
Infrastructure faults 0xC000005E or 0xC0000192 Route to operations, not the security queue

The stale credential row is analysis. It relies on the Caller Process Name and Workstation Name fields that Microsoft documents for the event [1].

When a failed logon pattern holds up, the next question is whether any of the failures ended in a success. Microsoft's 4625 page points to event 4624 as the matching successful logon event [1]. Search for 4624 events from the same source address and against the same account list after the failures.

MITRE lists three mitigations for password spraying: account lockout policies, multi-factor authentication, and password policies based on NIST guidelines [3]. It also warns that a lockout policy set too strictly can lock every account the attacker touched, which turns a spray into a denial of service [3].

Its account use mitigation adds conditional access, such as blocking logons from outside defined IP ranges or from anonymizing services [3]. Treat each of these as a containment option to test before the incident, not a default to switch on during one.

What did this guide check, and what is untested?

This guide reads Microsoft's 4625 documentation page (an archived Windows 10 edition, last updated 29 March 2026), MITRE ATT&CK's T1110.003 entry, the SigmaHQ rule, Microsoft's Get-WinEvent filtering guide and NIST's log management page.

No command, query or rule here was run against a live system. Field names follow Microsoft's sample event XML, and the Status table covers only the codes Microsoft's monitoring guidance lists, not every NTSTATUS value. The readings, the one-hour window and the threshold of ten are analysis.

What does Event ID 4625 monitoring change for your team this week?

Event ID 4625 gives a team an early, cheap signal that someone is guessing passwords, and the signal is only as good as the audit policy behind it. The first decision is coverage. Confirm that Audit Logon and Audit Account Lockout are on for every domain controller and every server that accepts outside connections, because an event that was never logged cannot be counted [1].

NIST's draft revision of SP 800-92 frames log management planning as the way to make sure you have the log data you need, and it is still a draft on NIST's project page [5].

The second decision is where the count happens. Per-account lockout thresholds miss a throttled spray, so the alert belongs on distinct accounts per source address, with an allowlist that has an owner.

The third is what 4625 cannot show. MITRE documents LDAP and Kerberos attempts that are less likely to trigger it [3], so 4625 monitoring is one input to authentication monitoring, not the whole of it. Pair it with success events from the same source, because a failed run followed by a success is the moment a case moves from noise to incident.

This week, run the Get-WinEvent query above on one domain controller and one jump server. Write down the top five source addresses by distinct accounts, mark the ones you expect, and open a ticket for each one nobody can explain.

Sources & References

  1. Microsoft Learn, 4625(F): An account failed to log on (Windows 10 archived documentation, updated 29 March 2026).
  2. Microsoft Learn, Creating Get-WinEvent queries with FilterHashtable (PowerShell documentation).
  3. MITRE ATT&CK, Brute Force: Password Spraying, T1110.003 (version 1.8, last modified 4 November 2026).
  4. SigmaHQ rule "Multiple Users Remotely Failing To Authenticate From Single Source" (status: unsupported), as displayed at Detection.FYI, which links to the rule file in the SigmaHQ repository.
  5. NIST Computer Security Resource Center, Log Management project (SP 800-92 and draft SP 800-92 Rev. 1).
NK

Cybersecurity Researcher & Technical Editor

UnpanicTech publishes independent cybersecurity research and defensive security analysis. Naseem Khan serves as Technical Editor, with a focus on vulnerability analysis, defensive security, incident response, cloud security, and practical security engineering.

Leave a Comment (Authenticated Users)