How the attack works
Password spraying turns guessing around. The attacker takes a list of the organization's addresses and tries one likely password against each of them, then a second password some time later. Every account sees only one or two failures, which stays under what an account owner or a per-account count would notice. Across hundreds of accounts a few usually have a weak or reused password. The addresses are easy to collect, because they follow a naming scheme.
Google Workspace and Cloud Identity accounts are worth spraying because the same identity opens mail, files and Google Cloud. Organizations that send their users to an identity provider for single sign-on move this exposure to the identity provider. Accounts that sign in with a Google password remain, and Google states that super administrators at admin.google.com always do.
The Login audit log records login_failure for each unsuccessful attempt, with login_type set to google_password for a Google account password. Organizations that share Workspace data with Google Cloud receive the event in Cloud Audit Logs at organization level as a Data Access entry of login.googleapis.com, with the account in principalEmail and the source in requestMetadata.callerIp, as in the example entry Google publishes. The rule on this page counts how many different accounts one source address failed for. An attacker who spreads the attempts over many addresses stays under it, which is the main limit of a rule of this kind.
What an attacker needs
- A list of valid account addresses, which can be derived from public names and the organization's naming scheme.
- Accounts that sign in with a Google password and have a common, reused or breached password. Accounts without 2-Step Verification turn a hit into access.
Impact
One or more valid passwords out of many accounts. Each account without 2-Step Verification is a session as that user, in Workspace and in Google Cloud. From the first mailbox the attacker reads the directory, sends mail that colleagues trust and looks for an account with more privileges. Spraying is often the first step of a longer intrusion, because it needs nothing but a list of names.
Detection rules
Failed Password Sign-Ins for Many Accounts From One Address in Google Workspace
Counts, per source address (requestMetadata.callerIp), the number of different accounts (principalEmail) of a Google Workspace or Cloud Identity organization for which a sign-in with a Google account password failed (Login audit event login_failure with the login_type google_password). Ten or more different accounts from one address within one hour fit password spraying, in which a few common passwords are tried against many accounts so that no single account sees many failures. Google's published example of a Login audit entry carries the account in authenticationInfo.principalEmail and the source in requestMetadata.callerIp. The rule finds the login type by searching the serialized event parameters in protoPayload.metadata for the value google_password. An attacker who spreads the attempts over many addresses stays under the count for each one. Users who sign in through an identity provider enter their password there, and a wrong password at the identity provider does not produce this event. The count and the window are starting points to tune. The entries reach Cloud Audit Logs as Data Access entries of login.googleapis.com at organization level only when Google Workspace data sharing with Google Cloud is enabled, and reach the SIEM only when the organization-level Data Access logs are routed to it.
Fires when one protoPayload.requestMetadata.callerIp has at least 10 different protoPayload.authenticationInfo.principalEmail values within 1 hour.
title: Failed Google Password Sign-In in Google Workspace (Spraying Base Event)
id: a3425038-0074-4188-bba7-346a6a29ee7d
name: workspace_failed_google_password_sign_in_many_accounts
status: experimental
description: >-
Base event of the threshold rule in this file. A Login audit entry for an
unsuccessful login attempt (login_failure) whose event parameters contain
google_password, the documented value of login_type for a Google account
password, as it arrives in Cloud Audit Logs under login.googleapis.com.
logsource:
product: gcp
service: gcp.audit
detection:
selection:
protoPayload.serviceName: login.googleapis.com
protoPayload.methodName: google.login.LoginService.loginFailure
protoPayload.metadata.event|contains: google_password
condition: selection
level: low
---
title: Failed Password Sign-Ins for Many Accounts From One Address in Google Workspace
id: 2e7c36c4-b8cf-4a74-a871-95c08f9c697a
status: experimental
description: >-
Counts, per source address (requestMetadata.callerIp), the number of
different accounts (principalEmail) of a Google Workspace or Cloud
Identity organization for which a sign-in with a Google account password
failed (Login audit event login_failure with the login_type
google_password). Ten or more different accounts from one address within
one hour fit password spraying, in which a few common passwords are tried
against many accounts so that no single account sees many failures.
Google's published example of a Login audit entry carries the account in
authenticationInfo.principalEmail and the source in
requestMetadata.callerIp. The rule finds the login type by searching the
serialized event parameters in protoPayload.metadata for the value
google_password. An attacker who spreads the attempts over many addresses
stays under the count for each one. Users who sign in through an identity
provider enter their password there, and a wrong password at the identity
provider does not produce this event. The count and the window are
starting points to tune. The entries reach Cloud Audit Logs as Data
Access entries of login.googleapis.com at organization level only when
Google Workspace data sharing with Google Cloud is enabled, and reach the
SIEM only when the organization-level Data Access logs are routed to it.
references:
- https://docs.cloud.google.com/logging/docs/audit/gsuite-audit-logging
- https://developers.google.com/workspace/admin/reports/v1/appendix/activity/login
- https://attack.mitre.org/techniques/T1110/003/
author: Forestall
date: 2026-11-18
tags:
- attack.credential-access
- attack.t1110.003
correlation:
type: value_count
rules:
- workspace_failed_google_password_sign_in_many_accounts
group-by:
- protoPayload.requestMetadata.callerIp
timespan: 1h
condition:
field: protoPayload.authenticationInfo.principalEmail
gte: 10
falsepositives:
- Many users behind one office, VPN or proxy address who fail in the same period, for example after a password expiry. Exclude your own egress addresses in a filter_* selection of the base rule or give them a higher count.
- A shared device or kiosk on which many people sign in and mistype their passwords.
- The values of 10 accounts and one hour are starting points, not tested limits. Measure how many accounts normally fail per address in your organization and adjust both.
level: medium
False positives
- Many users behind one office, VPN or proxy address who fail in the same period, for example after a password expiry. Exclude your own egress addresses in a filter_* selection of the base rule or give them a higher count.
- A shared device or kiosk on which many people sign in and mistype their passwords.
- The values of 10 accounts and one hour are starting points, not tested limits. Measure how many accounts normally fail per address in your organization and adjust both.
Detection logic
Patterns that need more than one event, a baseline or knowledge of your own environment. Each is a hypothesis to build and tune in your SIEM, not a ready rule.
Successful sign-in from an address that failed for many accounts
A spray matters when one of its attempts works. A successful sign-in from the address that just failed for many different accounts is very likely that hit, and it names the compromised account.
- An address in
requestMetadata.callerIpthat producedgoogle.login.LoginService.loginFailureentries for many differentauthenticationInfo.principalEmailvalues, as the threshold rule of this page counts them. - Within 24 hours before or after, a
google.login.LoginService.loginSuccessentry oflogin.googleapis.comwith the samecallerIp. - Raise the account in
authenticationInfo.principalEmailof the success with high priority, unless the address is one of your own egress addresses.
Eventsgoogle.login.LoginService.loginFailuregoogle.login.LoginService.loginSuccess
LimitsBehind an office or VPN address, failures of some users and successes of others are normal, so this needs your list of egress addresses. A sprayer who signs in to the hit account from a different address later is not found. The 24 hours are a starting point.
Triage
- Read the source address in requestMetadata.callerIp and find out what it is. Your own office, VPN or proxy address explains many users failing in the same period. A hosting provider or a country you have no staff in does not.
- List the accounts the address failed for (principalEmail). Addresses in alphabetical order, accounts of people who left, or a mix across every department point to a list being worked through.
- Look at the timing. The same small number of attempts per account, repeated at regular intervals, is how a spray paces itself.
- Search for google.login.LoginService.loginSuccess entries from the same address. Any success is a compromised account until shown otherwise.
- Check whether Google raised suspicious login events for accounts in the list, and whether the same address appears in sign-ins to other services you run.
Response
Once the attack is confirmed:
- For every account that signed in successfully from the spraying address, reset the password, reset the sign-in cookies and review the 2-Step Verification methods and backup codes.
- Review what those accounts did since the sign-in, in Workspace and in Google Cloud, including mail forwarding, app access and role bindings, and undo what was not the owner.
- Force a password change for the accounts that were sprayed and have no 2-Step Verification, since the attacker may come back to them with the next password.
- Enforce 2-Step Verification for the organizational units that were hit, and remove accounts that are no longer needed.
- Watch for the same pattern from new addresses, because a spray that was noticed usually moves and continues.
Mitigation
- Enforce 2-Step Verification for every user, with security keys or passkeys for administrators, so that a correct password is not enough to sign in.
- Require strong passwords in the password settings of the Admin console. Google's strength check refuses common passwords and passwords known to be compromised, which are exactly what a spray tries.
- Use single sign-on with your identity provider for users where you can, and apply its spray protection and lockout there.
- Find and remove dormant accounts and accounts that never enrolled in 2-Step Verification, because they are the ones a spray opens.
- Share Workspace data with Google Cloud and route the organization-level audit logs to the SIEM, including the Data Access log that carries the Login audit events.
Frequently asked questions
How does password spraying differ from password guessing in the logs?
Guessing produces many failed sign-ins for one account. Spraying produces one or two failed sign-ins for each of many accounts, so a count per account stays low. The rule on this page therefore counts the number of different accounts that one source address failed for. The page Password Guessing Against One Account in Google Workspace covers the other pattern.
Can this rule find a spray that comes from many addresses?
No. It groups by source address, so an attacker who uses a different address for every few attempts stays under the count for each one. For that case, count failed sign-ins for the whole organization per hour and compare with your normal level, and look for many accounts failing with the same pattern at the same time.
Why do my own office addresses trigger the rule?
Many users share one public address behind an office network, a VPN or a proxy. After a password expiry or a change of the sign-in flow, several of them can fail within the same hour. Exclude your own egress addresses from the rule or give them a higher count, and remember that an attacker inside your network would then need a different detection.
Does the event show which password was tried?
No. The audit entry names the account, the source address, the login type and the challenge methods. It never contains the password. Whether a spray used one password or several can only be inferred from the timing, for example one attempt per account in a first round and a second attempt per account an hour later.
References
- MITRE ATT&CK T1110.003: Brute Force: Password Spraying
- Audit logs for Google Workspace (Cloud Audit Logs, with an example Login audit entry)
- Configure audit logs for Google Workspace
- Login audit events (login_failure, login_type)
- Super administrator SSO
- Enforce and monitor password requirements for users
- Deploy 2-Step Verification