Detection rules › Splunk

Windows Suspicious Burst of Password Changes

Status
production
Severity
low
Group by
computer_name, dest, time_bucket, user
Author
Raven Tait, Splunk
Source
github.com/splunk/security_content

A regular user account performed rapid, repeated password changes across multiple local accounts within a 2-second window. This pattern is consistent with automated credential manipulation tools that cycle account passwords to deny access to defenders or escalate privileges The speed and volume of changes indicates scripted or tooled activity rather than manual administration, as legitimate password resets do not occur at machine speed across multiple accounts simultaneously.

Known false positives

  • Service accounts may be responsible for the creation, deletion or modification of accounts for legitimate purposes. Filter as needed.

MITRE ATT&CK coverage

TacticTechniques
Privilege Escalation

Telemetry coverage

Rule body

name: Windows Suspicious Burst of Password Changes
id: 0dcfcab9-e936-43bc-8203-005e17dc0744
version: 1
creation_date: '2026-04-29'
modification_date: '2026-04-29'
author: Raven Tait, Splunk
status: production
type: TTP
description: |-
    A regular user account performed rapid, repeated password changes across multiple local accounts within a 2-second window.
    This pattern is consistent with automated credential manipulation tools that cycle account passwords to deny access to defenders or escalate privileges The speed and volume of changes indicates scripted or tooled activity rather than manual administration, as legitimate password resets do not occur at machine speed across multiple accounts simultaneously.
data_source:
    - Windows Event Log Security 4723
    - Windows Event Log Security 4724
search: |-
    `wineventlog_security`
    (
        EventCode=4723
        OR
        EventCode=4724
    )
    | eval log_time = strptime(SystemTime, "%Y-%m-%dT%H:%M:%S.%6NZ")
    | eval log_time = coalesce(log_time, _time)
    | eval time_bucket = floor(log_time / 2) * 2
    | eval SubjectRID = tonumber(replace(SubjectUserSid, ".*-(\d+)$", "\1"))
    | where SubjectRID >= 1000
    | stats count AS EventsForAccount
            min(log_time) AS firstTime
            max(log_time) AS lastTime
      by Computer TargetUserName SubjectUserName time_bucket
    
    | where EventsForAccount >= 4
    | eval DurationSecs = round(lastTime - firstTime, 3), DurationSecs = if(DurationSecs = 0, 0.001, DurationSecs), ChangeRatePerSec = round(EventsForAccount / DurationSecs, 2)
    | rename Computer as dest
    
    | stats dc(TargetUserName) AS AffectedAccounts
            values(TargetUserName) AS Accounts
            values(SubjectUserName) AS InitiatingAccounts
            sum(EventsForAccount) AS TotalEvents
            max(ChangeRatePerSec) AS PeakChangeRatePerSec
            min(DurationSecs) AS FastestBurstSecs
            min(firstTime) AS firstTime
            max(lastTime) AS lastTime
      by dest time_bucket
    
    | `security_content_ctime(firstTime)`
    | `security_content_ctime(lastTime)`
    | `windows_suspicious_burst_of_password_changes_filter`
how_to_implement: |-
    To successfully implement this search, you need to be ingesting EventID 4723 and EventID 4724 events with the Windows TA.
known_false_positives: |-
    Service accounts may be responsible for the creation, deletion or modification of accounts for legitimate purposes. Filter as needed.
references:
    - https://github.com/Nightmare-Eclipse/BlueHammer
finding:
    title: Rapid burst of password changes of local accounts by $InitiatingAccounts$ on $dest$
    entity:
        field: dest
        type: system
        score: 20
analytic_story:
    - Windows Privilege Escalation
    - BlueHammer
asset_type: Endpoint
mitre_attack_id:
    - T1068
product:
    - Splunk Enterprise
    - Splunk Enterprise Security
    - Splunk Cloud
category: endpoint
security_domain: endpoint
cve:
    - CVE-2026-33825

Stages and Predicates

Stage 1: search

`wineventlog_security`
(
    EventCode=4723
    OR
    EventCode=4724
)

Stage 2: eval

| eval log_time = strptime(SystemTime, "%Y-%m-%dT%H:%M:%S.%6NZ")

Stage 3: eval

| eval log_time = coalesce(log_time, _time)

Stage 4: eval

| eval time_bucket = floor(log_time / 2) * 2

Stage 5: eval

| eval SubjectRID = tonumber(replace(SubjectUserSid, ".*-(\d+)$", "\1"))

Stage 6: where

| where SubjectRID >= 1000

Stage 7: stats

| stats count AS EventsForAccount
        min(log_time) AS firstTime
        max(log_time) AS lastTime
  by Computer TargetUserName SubjectUserName time_bucket

Stage 8: where

| where EventsForAccount >= 4

Stage 9: eval

| eval DurationSecs = round(lastTime - firstTime, 3), DurationSecs = if(DurationSecs = 0, 0.001, DurationSecs), ChangeRatePerSec = round(EventsForAccount / DurationSecs, 2)
DurationSecs =
ifDurationSecs = 00.001
elseDurationSecs

Stage 10: rename

| rename Computer as dest

Stage 11: stats

| stats dc(TargetUserName) AS AffectedAccounts
        values(TargetUserName) AS Accounts
        values(SubjectUserName) AS InitiatingAccounts
        sum(EventsForAccount) AS TotalEvents
        max(ChangeRatePerSec) AS PeakChangeRatePerSec
        min(DurationSecs) AS FastestBurstSecs
        min(firstTime) AS firstTime
        max(lastTime) AS lastTime
  by dest time_bucket

Stage 12: search

| `security_content_ctime(firstTime)`

Stage 13: search

| `security_content_ctime(lastTime)`

Stage 14: search

| `windows_suspicious_burst_of_password_changes_filter`

Indicators

These rows show field, operator, and value matches.