Detect Multi-Factor Authentication Fatigue (MFA Bombing) Attack in Sumo Logic CSE
MFA fatigue (also called MFA bombing or push flooding) is a social engineering technique where an attacker who has obtained valid credentials uses repeated MFA push notifications to wear down the victim into approving an authentication request out of annoyance or confusion. Scattered Spider pioneered this at scale, compromising MGM Resorts, Caesars Entertainment, and numerous UK-based organisations. The attacker sends dozens of Authenticator app push notifications in rapid succession, sometimes at 3am to catch sleeping victims, until one is approved. Some variants include calling the victim while bombing, claiming to be IT support (vishing), and guiding them to approve the 'legitimate' request. NCSC and CISA issued a joint advisory on this technique in 2023. With valid M365 credentials available from password spray or phishing, MFA fatigue is the primary way Scattered Spider bypasses MFA.
MITRE ATT&CK
- Tactic
- Credential Access
Sumo Detection Query
_sourceCategory=azure/signinlogs OR _sourceCategory=azure/aad/signin
| json field=_raw "properties.userPrincipalName" as user
| json field=_raw "properties.authenticationRequirement" as auth_req
| json field=_raw "properties.status.errorCode" as error_code
| json field=_raw "properties.ipAddress" as ip_address
| json field=_raw "properties.location" as location
| json field=_raw "properties.appDisplayName" as app_name
| where auth_req = "multiFactorAuthentication"
| where error_code in ("0", "50074", "50076", "500121", "500133")
| eval mfa_result = if(error_code = "0", "SUCCESS",
if(error_code = "500121", "DENIED",
if(error_code = "500133", "NO_RESPONSE", "PROMPTED")))
| timeslice 10m
| stats
count AS TotalPrompts,
countif(mfa_result = "SUCCESS") AS Approvals,
countif(mfa_result = "DENIED") AS Denials,
countif(mfa_result = "NO_RESPONSE") AS NoResponses,
dcount(ip_address) AS UniqueSourceIPs,
values(location) AS Locations,
values(app_name) AS Applications
by user, _timeslice
| where TotalPrompts >= 5
| eval FatigueSuccess = if(Approvals >= 1 AND (Denials + NoResponses) >= 3, "YES", "NO")
| eval Severity = if(FatigueSuccess = "YES", "CRITICAL", "HIGH")
| eval ThreatType = "MFA_Fatigue_PushBombing"
| eval ThreatActors = "Scattered Spider, Lapsus$, 0ktapus"
| fields user, _timeslice, TotalPrompts, Approvals, Denials, NoResponses, UniqueSourceIPs, FatigueSuccess, Severity, ThreatType, Locations, Applications
| sort by TotalPrompts descSumo Logic query ingesting Azure AD sign-in logs and parsing raw JSON payloads to extract user identity, authentication requirement, MFA outcome error codes, and source IP. Events are bucketed into 10-minute timeslices per user using the timeslice operator. The stats command counts MFA outcome categories. The CRITICAL/HIGH severity split on FatigueSuccess distinguishes confirmed victim capitulation (approval after 3+ failures) from high-volume prompting alone. The values() aggregation on location and app_name provides immediate enrichment context for triage without requiring a secondary lookup.
Data Sources
Required Tables
False Positives & Tuning
- Users with persistent MFA app synchronisation problems (clock skew on hardware tokens or soft tokens) who generate repeated failures before escalating to IT and resolving the configuration issue
- Third-party federated SSO integrations that silently re-authenticate against Azure AD during token refresh cycles, triggering MFA Conditional Access policies and accumulating timeout events without user awareness
- Bulk licence assignment or tenant migration scripts that programmatically trigger authentication flows for many accounts, some of which may have MFA configured inconsistently
Other platforms for THREAT-EntraID-MFAFatigue
Testing Methodology
Validate this detection against 1 adversary technique from Atomic Red Team. Each test below lists the behaviour to exercise and the telemetry you should expect to see. Executable commands and cleanup steps are available with Pro.
- Test 1 Simulate MFA Push Bombing via Microsoft Authentication Library
Expected signal: Azure AD Sign-in logs record 10 MFA prompts for the target user within a 5-minute window, each showing error code 50076 (MFA required) or 500133 (not completed).
Response Playbook
Triage
- Contact the affected user directly via phone (NOT via Teams or email — the account may be compromised). Ask if they received unusual MFA push notifications and whether they approved any.
- If the user confirms receiving unexpected MFA pushes, treat as active MFA fatigue attack in progress. Disable legacy auth and block the attacker IP immediately.
- Check Azure AD sign-in logs for the attacker's source IP: is this a known VPN exit node, Tor, or residential proxy? This helps attribute the attack vector.
- Determine if the user's password was obtained via a prior password spray (check for spray activity against this user's account in the preceding 24-48 hours).
- If MFA was ultimately approved: treat as compromised account. Check MailItemsAccessed, Teams messages, and SharePoint activity from the attacker's IP.
Containment
- Disable the user's account immediately while investigating if MFA approval was detected.
- Block the attacker IP in Entra ID Named Locations.
- Enable 'Number Matching' in Microsoft Authenticator if not already deployed — this requires the user to enter a number shown on the sign-in screen into the Authenticator app, defeating push bombing.
- Enable 'Additional context' in Microsoft Authenticator to show the application name and location on the MFA push — makes fake MFA requests obvious to users.
- Consider migrating from push notifications to FIDO2 security keys or certificate-based authentication — these cannot be fatigue-bombed.
Evidence Collection
- Azure AD Sign-in logs: all MFA events for the affected user over the incident window
- Microsoft Authenticator logs (if Intune-managed device): push notification delivery timestamps
- Entra ID Identity Protection risk events for the user
- Network logs from corporate proxy showing attacker IP activity
Escalation Criteria
- !MFA approval confirmed following 3+ denials — user likely approved out of fatigue
- !Simultaneous vishing call to user while push bombing is detected (caller claims to be IT helpdesk)
- !User account is a Global Admin, Security Admin, or privileged account
- !Post-approval activity: sign-in from attacker IP accessing SharePoint, Exchange, or Teams
Investigation Guide
Related Techniques
Forensic Artifacts
- >
Azure AD Sign-in logs for the user: MFA prompt/denial/approval sequence with source IPs - >
Microsoft Authenticator notification history (from Intune MDM policy if managed) - >
Entra ID audit logs: any MFA method changes after the fatigue event - >
Call logs: did user receive phone calls coinciding with push bombing (vishing component)? - >
O365 audit logs: post-compromise activity if MFA was approved
Tuning Guidance
The 5-prompt threshold in 10 minutes is appropriate for most SMBs. Users who use MFA regularly may occasionally trigger 2-3 prompts due to connectivity, but 5+ in 10 minutes is strongly anomalous. The 'fatigue success' detection (approval following denials) is the highest-fidelity alert and should be treated as critical. Consider enabling Microsoft Authenticator's 'Report Suspicious Activity' feature which automatically marks risky sign-ins and disables the user account — this provides automatic containment.
Hunting Queries
Hunt for users with high rates of unresponded MFA prompts over 30 days — may indicate ongoing low-slow MFA fatigue campaign or users who routinely ignore unexpected prompts (a risk indicator regardless).
AADSignInLogs
| where TimeGenerated > ago(30d)
| where Status.errorCode == 500133 // MFA not completed
| summarize
MFASkips=count(),
Days=dcount(bin(TimeGenerated, 1d)),
IPs=make_set(IPAddress)
by UserPrincipalName
| where MFASkips > 10
| sort by MFASkips descindex=azure sourcetype="azure:aad:signin" properties.status.error_code=500133
| stats count AS MFASkips, dc(bin(_time, 86400)) AS Days
BY properties.user_principal_name
| where MFASkips > 10
| sort - MFASkipsAtomic Red Team Tests
Simulates MFA fatigue by programmatically triggering multiple MFA push notification requests for an account with known credentials. This tests detection of rapid MFA prompt generation.
Command
python3 -c "
import requests, time
for i in range(10):
r = requests.post('https://login.microsoftonline.com/<TENANT>/oauth2/v2.0/token',
data={'client_id':'d3590ed6-52b3-4102-aeff-aad2292ab01c',
'grant_type':'password','username':'<USER>','password':'<KNOWN_PASSWORD>',
'scope':'openid'})
print(f'Attempt {i+1}: {r.status_code}')
time.sleep(30)
"Expected Telemetry
Azure AD Sign-in logs record 10 MFA prompts for the target user within a 5-minute window, each showing error code 50076 (MFA required) or 500133 (not completed).
Expected Detection
Alert fires at 5+ MFA prompts within 10 minutes for the same user.