top of page

Com Vishing: Fake Help Desk Calls and Rogue MFA Enrollment

Writer: Inception Security
Inception Security
2 hours ago
17 min read
Bright office comic. At the Riverdale Community Office desk, a receptionist in a white shirt answers the phone, looking uneasy, with a speech bubble that says Hi, it's IT, I just need you to approve this. A laptop shows a green APPROVE button. In the center, a hand holds a phone labeled Pixel 10a that reads New MFA method added, hanging from an MFA key rack. A man in a light blue shirt writes on a whiteboard: new MFA method, minutes after sign-in, next to a sticky note that says hang up and call back. The caption reads Your help desk is the voice they borrow. Inception Security is in the corner.

What this means for your clients: Criminals are phoning staff while pretending to be the company's own IT support, talking them into signing in on a fake page, then quietly adding their own phone as a second login key and copying files. Agree on one way your real IT team proves who it is, and report any call that asks you to sign in or approve a prompt.


The caller says they are from IT and will help fix a sign-in problem. You open their page, type your password and approve the prompt. That page was a relay, and the caller now holds your session.


That is Com vishing as described in Unit 42's "Dialing for Data", published November 20, which covers vishing waves by Com-affiliated threat actors that led to data theft and data extortion attempts against financial services, healthcare and legal organizations, mostly in the U.S., UK and Australia. There is no software flaw to patch here. Without encryption, there may be no ransomware alarm; the first sign may be an extortion message in Microsoft Teams.


How Com vishing works, per Unit 42


Unit 42 attributes the broad activity to Bling Libra (aka ShinyHunters) and to clusters it tracks as CL-CRI-1116, CL-CRI-1147, CL-CRI-1182 and CL-CRI-1192, plus two new ones, CL-CRI-1264 (Vortex) and CL-CRI-1274 (Cinder), which overlap with activity Google tracks as UNC6671.


The callers pose as IT staff resolving an authentication issue and send the victim to a phishing page tailored to their organization, often a short-lived subdomain such as {company}.mfaregister[.]com. Some crews share Doko's Panel, an AiTM phishing kit with Okta-targeting variants. In one financial services case, Unit 42 recorded a full sequence from credential entry to an MFA number-match prompt, indicating a likely successful session capture. In our judgment, number matching alone will not stop a relay when the attacker is coaching the user on the phone.


Vortex, mostly hitting healthcare, pharmaceutical and medical organizations, enrolled an unauthorized MFA device within minutes of authenticating, automated exfiltration from Microsoft 365 mail, OneDrive and SharePoint with Python scripts (a python-httpx user agent), then sent Teams messages to employees pointing to a Tor-hosted negotiation site. Cinder replayed a persistent "keep me signed in" token to avoid Duo re-prompts, collected data with python-requests and the Microsoft Graph SDK (Unit 42's IOC list adds kiota-dotnet), and sent extortion by email and Teams from a compromised account. In a CL-CRI-1192 case, Unit 42 found a sideloaded browser extension named "Google Chrome" that logs keystrokes and steals auth cookies, assessed with moderate confidence as that cluster's build.


Unit 42 describes the MFA device names Pixel 10a and Pixel 9 Pro XL as "standard phone models, not standalone indicators of compromise."


Why MSPs should care


In Com vishing, your help desk is the voice they borrow. Users at small legal, healthcare and finance firms trust a call from IT, and for many of them IT is their MSP. MSPs that never told users how their techs prove who they are have left that trust open.


Start with three changes. Set a callback rule: our techs never call and send you a link to sign in or set up MFA; hang up and call us on the number you already have. Review every new MFA registration, and treat "I'm traveling, add my new phone" as high risk. Move admins and high-value users to phishing-resistant MFA, FIDO2 keys or device-bound passkeys. Unit 42 also recommends tighter help desk verification, extension allowlists, identity visibility and alerts on unauthorized data exports.


How to hunt


We have not run these in a live tenant; tune field names and test before trusting a clean result. Splunk assumes the Microsoft Cloud Services and Office 365 add-ons plus Sysmon, Elastic the Azure and Microsoft 365 integrations plus Defend, and KQL Sentinel plus Defender tables. The CrowdStrike Falcon LogScale identity hunts need you to ingest Entra ID and M365 logs into LogScale or Next-Gen SIEM. Every YOUR_ name, #repo and LogScale cloud field is a placeholder.


Test 1: every MFA registration in 30 days


This lists every security info registration in 30 days, by Entra's documented activity names, plus a summary row. No audit events means the source is missing, audit data starting inside the window means short coverage, and a quiet tenant points to the Entra registration report; none of them is clean.


Field mapping: Splunk reads azure:monitor:aad properties, Elastic reads azure.auditlogs fields, LogScale uses placeholder parser fields, and KQL reads AuditLogs.


False positives: onboarding and phone upgrades.


Splunk


index=YOUR_ENTRA_INDEX sourcetype="azure:monitor:aad" category=AuditLogs operationName IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info") earliest=-30d latest=now
| eval user=lower(coalesce(mvindex('properties.targetResources{}.userPrincipalName', 0), 'properties.initiatedBy.user.userPrincipalName')), ip='properties.initiatedBy.user.ipAddress', detail=mvjoin('properties.additionalDetails{}.value', " ")
| table _time user operationName ip detail
| append [search index=YOUR_ENTRA_INDEX sourcetype="azure:monitor:aad" category=AuditLogs earliest=-30d latest=now
    | stats count AS audit_events min(_time) AS earliest_audit count(eval(in(operationName, "User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"))) AS registrations
    | eval user="SUMMARY", operationName=case(audit_events=0, "1 SOURCE MISSING: no Entra audit events in 30 days", earliest_audit>relative_time(now(), "-29d"), "2 COVERAGE SHORT: audit data starts inside the 30 days", registrations=0, "3 no registrations in 30 days: confirm in the Entra registration report", true(), "4 source present: review every registration"), detail="audit events: ".audit_events.", registrations: ".registrations, _time=now()
    | table _time user operationName detail]
| sort - _time

Elastic


FROM logs-azure.auditlogs-*
| WHERE @timestamp >= NOW() - 30 days
| EVAL op = azure.auditlogs.operation_name
| EVAL reg = CASE(op IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"), 1, 0)
| EVAL detail = CASE(reg == 1, CONCAT(TO_STRING(@timestamp), " | ", COALESCE(`azure.auditlogs.properties.target_resources.0.user_principal_name`, azure.auditlogs.properties.initiated_by.user.userPrincipalName, "no user"), " | ", op, " | ", COALESCE(TO_STRING(source.ip), "no ip")), NULL)
| STATS audit_events = COUNT(*), earliest_audit = MIN(@timestamp), registrations = SUM(reg), rows = VALUES(detail)
| EVAL status = CASE(audit_events == 0, "1 SOURCE MISSING: no Entra audit events in 30 days", earliest_audit > NOW() - 29 days, "2 COVERAGE SHORT: audit data starts inside the 30 days", COALESCE(registrations, 0) == 0, "3 no registrations in 30 days: confirm in the Entra registration report", "4 source present: review every registration")

CrowdStrike


// needs Entra ID audit logs ingested into LogScale or Next-Gen SIEM; #repo and every field name here are placeholders for your parser
// set the time picker to the last 30 days
#repo=YOUR_ENTRA_REPO category=AuditLogs
| now()
| case {
    in(field=operationName, values=["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]) | reg := 1 | when := formatTime("%Y-%m-%d %H:%M", field=@timestamp, timezone="UTC") | detail := format("%s | %s | %s | %s", field=[when, user_principal_name, operationName, source_ip]) ;
    * | reg := 0 }
| stats([count(as=audit_events), min(@timestamp, as=earliest_audit), sum(reg, as=registrations), max(_now, as=now_ms), collect([detail])])
| default(field=registrations, value=0)
| case {
    test(audit_events == 0) | status := "1 SOURCE MISSING: no Entra audit events in 30 days" ;
    test(earliest_audit > now_ms - 2505600000) | status := "2 COVERAGE SHORT: audit data starts inside the 30 days" ;
    test(registrations == 0) | status := "3 no registrations in 30 days: confirm in the Entra registration report" ;
    * | status := "4 source present: review every registration" }

Microsoft KQL


let ops = dynamic(["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]);
let audit = AuditLogs | where TimeGenerated > ago(30d);
let regs = audit
    | where OperationName in (ops)
    | extend User = tolower(coalesce(tostring(TargetResources[0].userPrincipalName), tostring(InitiatedBy.user.userPrincipalName))), IP = tostring(InitiatedBy.user.ipAddress)
    | project TimeGenerated, User, Activity = OperationName, IP, Detail = tostring(AdditionalDetails);
let health = audit
    | summarize AuditEvents = count(), EarliestAudit = min(TimeGenerated), Registrations = countif(OperationName in (ops))
    | extend Activity = case(AuditEvents == 0, "1 SOURCE MISSING: no Entra audit events in 30 days", EarliestAudit > ago(29d), "2 COVERAGE SHORT: audit data starts inside the 30 days", Registrations == 0, "3 no registrations in 30 days: confirm in the Entra registration report", "4 source present: review every registration")
    | project TimeGenerated = now(), User = "SUMMARY", Activity, IP = "", Detail = strcat("audit events: ", AuditEvents, ", registrations: ", Registrations);
union health, regs
| order by TimeGenerated desc

Hunt 1: new MFA method right after a new IP


This flags a registration from the Unit 42 IP, from an IP with no earlier successful sign-in in the 60-day lookback, or within 60 minutes of the first sign-in from that IP, our window, plus any registration from an IP with a medium or high risk sign-in.


Field mapping: Splunk, LogScale and KQL include non-interactive sign-ins, because Cinder replayed tokens; every query matches on user and IP.


False positives: new hires and travel. A registration from an established IP may not flag here; Test 1 still lists it.


Splunk


index=YOUR_ENTRA_INDEX sourcetype="azure:monitor:aad" earliest=-60d latest=now (category=SignInLogs OR category=NonInteractiveUserSignInLogs OR (category=AuditLogs operationName IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info")))
| eval kind=if(category="AuditLogs", "reg", "signin")
| eval user=lower(coalesce('properties.userPrincipalName', mvindex('properties.targetResources{}.userPrincipalName', 0), 'properties.initiatedBy.user.userPrincipalName')), ip=coalesce('properties.ipAddress', 'properties.initiatedBy.user.ipAddress')
| eval ok=if(kind="signin" AND 'properties.status.errorCode'=0, 1, 0), risky=if(ok=1 AND in(lower('properties.riskLevelDuringSignIn'), "medium", "high"), 1, 0)
| where ok=1 OR (kind="reg" AND _time>=relative_time(now(), "-30d"))
| stats min(eval(if(ok=1, _time, null()))) AS first_ok max(risky) AS risky min(eval(if(kind="reg", _time, null()))) AS reg_time values(eval(if(kind="reg", operationName, null()))) AS activities BY user ip
| where isnotnull(reg_time)
| eval seconds_after=reg_time-first_ok, minutes_after=round(seconds_after/60, 2)
| eval flag=case(ip="193.149.176.62", "1 Unit 42 IP 193.149.176.62", isnull(first_ok) OR seconds_after<0, "2 registration from an IP with no earlier sign-in in the lookback", seconds_after<=3600, "3 first sign-in from this IP, then registration within 60 minutes", risky=1, "4 risky sign-in from this IP", true(), null())
| where isnotnull(flag)
| sort flag
| convert ctime(first_ok) ctime(reg_time)

Elastic


FROM logs-azure.signinlogs-*, logs-azure.auditlogs-*
| WHERE @timestamp >= NOW() - 60 days
| EVAL is_reg = CASE(azure.auditlogs.operation_name IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info") AND @timestamp >= NOW() - 30 days, 1, 0)
| EVAL ok = CASE(azure.signinlogs.properties.status.error_code == 0, 1, 0)
| WHERE is_reg == 1 OR ok == 1
| EVAL user = TO_LOWER(COALESCE(azure.signinlogs.properties.user_principal_name, `azure.auditlogs.properties.target_resources.0.user_principal_name`, azure.auditlogs.properties.initiated_by.user.userPrincipalName)), ip = TO_STRING(source.ip)
| EVAL ok_time = CASE(ok == 1, @timestamp, NULL), reg_time = CASE(is_reg == 1, @timestamp, NULL), risky = CASE(ok == 1 AND TO_LOWER(azure.signinlogs.properties.risk_level_during_signin) IN ("medium", "high"), 1, 0)
| STATS first_ok = MIN(ok_time), reg_time = MIN(reg_time), risky = MAX(risky), activities = VALUES(azure.auditlogs.operation_name) BY user, ip
| WHERE reg_time IS NOT NULL
| EVAL seconds_after = DATE_DIFF("second", first_ok, reg_time)
| EVAL flag = CASE(ip == "193.149.176.62", "1 Unit 42 IP 193.149.176.62", first_ok IS NULL OR reg_time < first_ok, "2 registration from an IP with no earlier sign-in in the lookback", reg_time <= first_ok + 60 minutes, "3 first sign-in from this IP, then registration within 60 minutes", risky == 1, "4 risky sign-in from this IP", NULL)
| WHERE flag IS NOT NULL
| SORT flag

CrowdStrike


// needs Entra ID sign-in and audit logs ingested into LogScale or Next-Gen SIEM; #repo and every field name here are placeholders for your parser
// set the time picker to the last 60 days (30 days of registrations plus 30 days of sign-in history, our choice)
#repo=YOUR_ENTRA_REPO
| now()
| case {
    category=AuditLogs | in(field=operationName, values=["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]) | test(@timestamp > _now - 2592000000) | reg_time := @timestamp ;
    category=/SignInLogs$/ error_code="0" risk_level=/^(medium|high)$/i | ok_time := @timestamp | risky := 1 ;
    category=/SignInLogs$/ error_code="0" | ok_time := @timestamp | risky := 0 }
| groupBy([user_principal_name, source_ip], function=[min(ok_time, as=first_ok), min(reg_time, as=reg_time), max(risky, as=risky), collect([operationName])])
| default(field=reg_time, value=0)
| default(field=first_ok, value=0)
| default(field=risky, value=0)
| test(reg_time > 0)
| seconds_after := (reg_time - first_ok) / 1000
| case {
    source_ip="193.149.176.62" | flag := "1 Unit 42 IP 193.149.176.62" ;
    test(first_ok == 0) | flag := "2 registration from an IP with no earlier sign-in in the lookback" ;
    test(seconds_after < 0) | flag := "2 registration from an IP with no earlier sign-in in the lookback" ;
    test(seconds_after <= 3600) | flag := "3 first sign-in from this IP, then registration within 60 minutes" ;
    test(risky == 1) | flag := "4 risky sign-in from this IP" }
| sort(flag, order=asc)

Microsoft KQL


let ops = dynamic(["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]);
let signins = union SigninLogs, AADNonInteractiveUserSignInLogs
    | where TimeGenerated > ago(60d) and (ResultType == "0" or ResultType =~ "Success")
    | extend User = tolower(UserPrincipalName), IP = IPAddress, Risky = iff(RiskLevelDuringSignIn in ("medium", "high"), 1, 0)
    | summarize FirstOk = min(TimeGenerated), Risky = max(Risky) by User, IP;
AuditLogs
| where TimeGenerated > ago(30d) and OperationName in (ops)
| extend User = tolower(coalesce(tostring(TargetResources[0].userPrincipalName), tostring(InitiatedBy.user.userPrincipalName))), IP = tostring(InitiatedBy.user.ipAddress)
| summarize RegTime = min(TimeGenerated), Activities = make_set(OperationName) by User, IP
| join kind=leftouter signins on User, IP
| extend MinutesAfter = datetime_diff("minute", RegTime, FirstOk)
| extend Flag = case(IP == "193.149.176.62", "1 Unit 42 IP 193.149.176.62", isnull(FirstOk) or RegTime < FirstOk, "2 registration from an IP with no earlier sign-in in the lookback", RegTime <= FirstOk + 60m, "3 first sign-in from this IP, then registration within 60 minutes", Risky == 1, "4 risky sign-in from this IP", "")
| where isnotempty(Flag)
| project User, IP, Flag, RegTime, FirstOk, MinutesAfter, Activities
| order by Flag asc

Hunt 2: Graph and Python agents pulling data


This flags matching user agents on a user's Graph requests or audited file reads, by day, with 100 events a day as our volume line. python-httpx matches by prefix because Unit 42 gives no version.


Field mapping: Graph rows need Microsoft Graph activity logs on in diagnostic settings, with an object ID as UserId; file rows come from the Unified Audit Log, where MailItemsAccessed has no user agent.


False positives: internal scripts and apps use the Graph SDK; the signal is a person's account suddenly scripting from a new IP.


Splunk


earliest=-30d latest=now ((index=YOUR_O365_INDEX sourcetype="o365:management:activity" Operation IN ("FileDownloaded", "FileAccessed", "FileSyncDownloadedFull", "MailItemsAccessed")) OR (index=YOUR_GRAPH_INDEX sourcetype=YOUR_GRAPH_ACTIVITY_SOURCETYPE))
| eval user=lower(coalesce(UserId, 'properties.userId')), ip=coalesce(ClientIP, 'properties.ipAddress'), ua=coalesce(UserAgent, 'properties.userAgent', ""), item=coalesce(ObjectId, 'properties.requestUri'), src=if(isnotnull(Operation), "UAL ".Operation, "Graph")
| where match(ua, "(?i)python-requests/2[.]34[.]2|python-httpx|kiota-dotnet/1[.]17[.]1|Microsoft[.]Graph[.]Client/6[.]2[.]0") AND isnotnull(user) AND user!=""
| bin _time span=1d
| stats count AS events dc(item) AS items values(ip) AS ips values(ua) AS agents values(src) AS sources BY _time user
| eval flag=if(events>=100, "1 volume: 100 or more events in a day", "2 scripted agent on a user account")
| sort flag - events

Elastic


FROM logs-o365.audit-*, logs-azure.graphactivitylogs-*
| WHERE @timestamp >= NOW() - 30 days
| EVAL ua = TO_LOWER(COALESCE(user_agent.original, azure.graphactivitylogs.properties.user_agent, "")), user = TO_LOWER(COALESCE(o365.audit.UserId, azure.graphactivitylogs.properties.user_id)), item = COALESCE(o365.audit.ObjectId, azure.graphactivitylogs.properties.request_uri), src = COALESCE(o365.audit.Operation, "Graph")
| WHERE ua RLIKE """.*(python-requests/2\.34\.2|python-httpx|kiota-dotnet/1\.17\.1|microsoft\.graph\.client/6\.2\.0).*""" AND user IS NOT NULL AND (src == "Graph" OR src IN ("FileDownloaded", "FileAccessed", "FileSyncDownloadedFull", "MailItemsAccessed"))
| EVAL day = DATE_TRUNC(1 day, @timestamp)
| STATS events = COUNT(*), items = COUNT_DISTINCT(item), ips = VALUES(source.ip), agents = VALUES(ua), sources = VALUES(src) BY day, user
| EVAL flag = CASE(events >= 100, "1 volume: 100 or more events in a day", "2 scripted agent on a user account")
| SORT flag, events DESC

CrowdStrike


// needs M365 audit and Graph activity logs ingested into LogScale or Next-Gen SIEM; #repo and every field name here are placeholders for your parser
// set the time picker to the last 30 days
#repo=YOUR_M365_REPO
| user_agent=/python-requests\/2\.34\.2|python-httpx|kiota-dotnet\/1\.17\.1|Microsoft\.Graph\.Client\/6\.2\.0/i
| user_id=*
| case { log_source=graph ; operation=/^(FileDownloaded|FileAccessed|FileSyncDownloadedFull|MailItemsAccessed)$/ }
| day := formatTime("%Y-%m-%d", field=@timestamp, timezone="UTC")
| groupBy([user_id, day], function=[count(as=events), count(item, distinct=true, as=items), collect([source_ip, user_agent, operation])])
| case { test(events >= 100) | flag := "1 volume: 100 or more events in a day" ; * | flag := "2 scripted agent on a user account" }
| sort(events, order=desc)

Microsoft KQL


let uas = @"(?i)python-requests/2\.34\.2|python-httpx|kiota-dotnet/1\.17\.1|Microsoft\.Graph\.Client/6\.2\.0";
union
    (MicrosoftGraphActivityLogs
    | where TimeGenerated > ago(30d) and isnotempty(UserId) and UserAgent matches regex uas
    | project TimeGenerated, User = UserId, IP = IPAddress, UserAgent, Source = "Graph", Item = RequestUri),
    (OfficeActivity
    | where TimeGenerated > ago(30d) and Operation in ("FileDownloaded", "FileAccessed", "FileSyncDownloadedFull", "MailItemsAccessed") and UserAgent matches regex uas
    | project TimeGenerated, User = tolower(UserId), IP = ClientIP, UserAgent, Source = strcat("UAL ", Operation), Item = OfficeObjectId)
| summarize Events = count(), Items = dcount(Item), IPs = make_set(IP, 20), Agents = make_set(UserAgent, 10), Sources = make_set(Source, 10) by User, Day = bin(TimeGenerated, 1d)
| extend Flag = iff(Events >= 100, "1 volume: 100 or more events in a day", "2 scripted agent on a user account")
| order by Flag asc, Events desc

Hunt 3: Unit 42 domains and IPs


This matches DNS, proxy, network and sign-in events against the 56 Unit 42 domains and subdomains, the connect-pages Cloudflare Pages lure and both IPs.


Field mapping: Splunk reads Sysmon EID 3 and 22 and CIM proxy fields, Elastic reads Defend DNS, Falcon reads DnsRequest and NetworkConnectIP4, and KQL reads Defender network events.


False positives: security tools resolving threat feeds. In our judgment, a keyword hunt on Unit 42's Table 2 words stays manual: words like sso and portal fill legitimate domains, so review first-seen domains carrying your company name.


Splunk


earliest=-30d latest=now ((index=YOUR_SYSMON_INDEX (EventCode=3 OR EventCode=22)) OR index=YOUR_PROXY_INDEX OR (index=YOUR_ENTRA_INDEX sourcetype="azure:monitor:aad" (category=SignInLogs OR category=NonInteractiveUserSignInLogs)))
| eval d=lower(coalesce(QueryName, url_domain, "")), ip=coalesce(DestinationIp, dest_ip, 'properties.ipAddress', "")
| eval dhit=if(match(d, "(^|[.])(add-mfa|login-options|apply-passkey|auth-sso|authenroll|configkeys|configurekey|editmysso|enableoskey|enrollmfa|enrollmfamethods|enrollssopasskey|fastpasskeys|hostsso|installpasskey|installsso|itkeyenroll|itkeyportal|itkeyupdate|keyconfiguration|keyenrollos|keyosregister|mfa-options|mfa-passkey|mfa-registry|mfaoptions|mfapasskeysetup|mfaregister|mfasettings|my-passkey|my-passkeys|myaccountapps|myaccountoptions|myapps2fa|mypasskeyapp|myssoapps|myssoauth|myssosettings|new-passkey|nowsso|onboardpasskey|oskeysetup|oursso|passkey-mfa|passkeyconnect|register-mfa|register-passkeys|registermymfa|registersso|settingsportal|signinoptions|ssokeyportal|ssosettings|startmypasskey|configuresso|makepasskey)[.]com$|^connect-pages-[a-z0-9-]+[.]pages[.]dev$"), d, null()), ihit=if(in(ip, "31.42.184.213", "193.149.176.62"), ip, null())
| where isnotnull(dhit) OR isnotnull(ihit)
| eval indicator=coalesce(dhit, ihit), who=coalesce('properties.userPrincipalName', user, src, host)
| stats count AS events min(_time) AS first_seen max(_time) AS last_seen values(sourcetype) AS seen_in BY indicator who
| sort - events
| convert ctime(first_seen) ctime(last_seen)

Elastic


FROM logs-endpoint.events.network-*, logs-azure.signinlogs-*
| WHERE @timestamp >= NOW() - 30 days
| EVAL d = TO_LOWER(COALESCE(dns.question.name, "")), ip = TO_STRING(COALESCE(destination.ip, source.ip))
| WHERE d RLIKE """(.*\.)?(add-mfa|login-options|apply-passkey|auth-sso|authenroll|configkeys|configurekey|editmysso|enableoskey|enrollmfa|enrollmfamethods|enrollssopasskey|fastpasskeys|hostsso|installpasskey|installsso|itkeyenroll|itkeyportal|itkeyupdate|keyconfiguration|keyenrollos|keyosregister|mfa-options|mfa-passkey|mfa-registry|mfaoptions|mfapasskeysetup|mfaregister|mfasettings|my-passkey|my-passkeys|myaccountapps|myaccountoptions|myapps2fa|mypasskeyapp|myssoapps|myssoauth|myssosettings|new-passkey|nowsso|onboardpasskey|oskeysetup|oursso|passkey-mfa|passkeyconnect|register-mfa|register-passkeys|registermymfa|registersso|settingsportal|signinoptions|ssokeyportal|ssosettings|startmypasskey|configuresso|makepasskey)\.com|connect-pages-[a-z0-9-]+\.pages\.dev""" OR ip IN ("31.42.184.213", "193.149.176.62")
| EVAL indicator = CASE(ip IN ("31.42.184.213", "193.149.176.62"), ip, d), who = COALESCE(host.name, azure.signinlogs.properties.user_principal_name)
| STATS events = COUNT(*), first_seen = MIN(@timestamp), last_seen = MAX(@timestamp), seen_in = VALUES(event.dataset) BY indicator, who
| SORT events DESC

CrowdStrike


// set the time picker to the last 30 days; this is endpoint DNS and network telemetry only
// checking Entra sign-ins from 193.149.176.62 needs Entra ID logs ingested (Hunt 1 flags that IP)
#event_simpleName=/^(DnsRequest|NetworkConnectIP4)$/
| case {
    DomainName=/(^|\.)(add-mfa|login-options|apply-passkey|auth-sso|authenroll|configkeys|configurekey|editmysso|enableoskey|enrollmfa|enrollmfamethods|enrollssopasskey|fastpasskeys|hostsso|installpasskey|installsso|itkeyenroll|itkeyportal|itkeyupdate|keyconfiguration|keyenrollos|keyosregister|mfa-options|mfa-passkey|mfa-registry|mfaoptions|mfapasskeysetup|mfaregister|mfasettings|my-passkey|my-passkeys|myaccountapps|myaccountoptions|myapps2fa|mypasskeyapp|myssoapps|myssoauth|myssosettings|new-passkey|nowsso|onboardpasskey|oskeysetup|oursso|passkey-mfa|passkeyconnect|register-mfa|register-passkeys|registermymfa|registersso|settingsportal|signinoptions|ssokeyportal|ssosettings|startmypasskey|configuresso|makepasskey)\.com$|^connect-pages-[a-z0-9-]+\.pages\.dev$/i | indicator := lower(DomainName) ;
    in(field=RemoteAddressIP4, values=["31.42.184.213", "193.149.176.62"]) | indicator := RemoteAddressIP4 }
| groupBy([indicator, aid], function=[count(as=events), min(@timestamp, as=first_seen), max(@timestamp, as=last_seen), collect([#event_simpleName])])
| sort(events, order=desc)

Microsoft KQL


let dom = @"(?i)(^|\.)(add-mfa|login-options|apply-passkey|auth-sso|authenroll|configkeys|configurekey|editmysso|enableoskey|enrollmfa|enrollmfamethods|enrollssopasskey|fastpasskeys|hostsso|installpasskey|installsso|itkeyenroll|itkeyportal|itkeyupdate|keyconfiguration|keyenrollos|keyosregister|mfa-options|mfa-passkey|mfa-registry|mfaoptions|mfapasskeysetup|mfaregister|mfasettings|my-passkey|my-passkeys|myaccountapps|myaccountoptions|myapps2fa|mypasskeyapp|myssoapps|myssoauth|myssosettings|new-passkey|nowsso|onboardpasskey|oskeysetup|oursso|passkey-mfa|passkeyconnect|register-mfa|register-passkeys|registermymfa|registersso|settingsportal|signinoptions|ssokeyportal|ssosettings|startmypasskey|configuresso|makepasskey)\.com$|^connect-pages-[a-z0-9-]+\.pages\.dev$";
let ips = dynamic(["31.42.184.213", "193.149.176.62"]);
union
    (union SigninLogs, AADNonInteractiveUserSignInLogs
    | where TimeGenerated > ago(30d) and IPAddress in (ips)
    | project TimeGenerated, Seen = "Entra sign-in", Who = UserPrincipalName, Indicator = IPAddress),
    (DeviceNetworkEvents
    | where Timestamp > ago(30d)
    | extend Host = extract(@"^(?:[a-z]+://)?([^/:]+)", 1, tolower(RemoteUrl))
    | where Host matches regex dom or RemoteIP in (ips)
    | project TimeGenerated = Timestamp, Seen = strcat("Defender ", ActionType), Who = DeviceName, Indicator = iff(RemoteIP in (ips), RemoteIP, Host))
| summarize Events = count(), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated), Seen = make_set(Seen, 10) by Indicator, Who
| order by Events desc

Hunt 4: the fake "Google Chrome" extension


This looks for the two published hashes, the folderextension file names and lookups of oskeysetup[.]com. In extension inventory, look by hand for a "Google Chrome" extension not from the store.


Field mapping: Sysmon EID 15 carries hashes; Falcon and Defender may not log every .crx or .zip write.


False positives: unlikely on the hashes.


Splunk


index=YOUR_SYSMON_INDEX (EventCode=11 OR EventCode=15 OR EventCode=22) earliest=-30d latest=now
| eval f=lower(replace(coalesce(TargetFilename, ""), "^.*[\\\\/]", "")), h=upper(coalesce(Hashes, Hash, "")), q=lower(coalesce(QueryName, ""))
| eval hit=case(match(h, "930898E8B65E9CF4E2595AC4092042A707CCA8922751E723AD8B86260655154A|EA8D30C42E758559E78BD1FE97F8849011BEB7F24C7F092A066A5A62FA7FBE6F"), "1 Unit 42 hash", in(f, "folderextension.crx", "folderextension.zip"), "2 Unit 42 file name", match(q, "(^|[.])oskeysetup[.]com$"), "3 oskeysetup.com lookup or download", true(), null())
| where isnotnull(hit)
| stats count AS events min(_time) AS first_seen max(_time) AS last_seen values(TargetFilename) AS files values(Image) AS processes BY host hit
| sort hit
| convert ctime(first_seen) ctime(last_seen)

Elastic


FROM logs-endpoint.events.file-*, logs-endpoint.events.network-*
| WHERE @timestamp >= NOW() - 30 days
| EVAL fname = TO_LOWER(COALESCE(file.name, "")), sha = TO_LOWER(COALESCE(file.hash.sha256, "")), q = TO_LOWER(COALESCE(dns.question.name, ""))
| EVAL hit = CASE(sha IN ("930898e8b65e9cf4e2595ac4092042a707cca8922751e723ad8b86260655154a", "ea8d30c42e758559e78bd1fe97f8849011beb7f24c7f092a066a5a62fa7fbe6f"), "1 Unit 42 hash", fname IN ("folderextension.crx", "folderextension.zip"), "2 Unit 42 file name", q RLIKE """(.*\.)?oskeysetup\.com""", "3 oskeysetup.com lookup or download", NULL)
| WHERE hit IS NOT NULL
| STATS events = COUNT(*), first_seen = MIN(@timestamp), last_seen = MAX(@timestamp), paths = VALUES(file.path), processes = VALUES(process.name) BY host.name, hit
| SORT hit

CrowdStrike


// set the time picker to the last 30 days
#event_simpleName=/Written$|^DnsRequest$/
| case {
    SHA256HashData=/^(930898e8b65e9cf4e2595ac4092042a707cca8922751e723ad8b86260655154a|ea8d30c42e758559e78bd1fe97f8849011beb7f24c7f092a066a5a62fa7fbe6f)$/i | hit := "1 Unit 42 hash" ;
    TargetFileName=/[\\\/]folderextension\.(crx|zip)$/i | hit := "2 Unit 42 file name" ;
    DomainName=/(^|\.)oskeysetup\.com$/i | hit := "3 oskeysetup.com lookup or download" }
| groupBy([aid, hit], function=[count(as=events), min(@timestamp, as=first_seen), max(@timestamp, as=last_seen), collect([TargetFileName])])
| sort(hit, order=asc)

Microsoft KQL


let hashes = dynamic(["930898e8b65e9cf4e2595ac4092042a707cca8922751e723ad8b86260655154a", "ea8d30c42e758559e78bd1fe97f8849011beb7f24c7f092a066a5a62fa7fbe6f"]);
union
    (DeviceFileEvents
    | where Timestamp > ago(30d)
    | where SHA256 in (hashes) or FileName in~ ("folderextension.crx", "folderextension.zip") or FileOriginUrl contains "oskeysetup.com"
    | extend Hit = case(SHA256 in (hashes), "1 Unit 42 hash", FileName in~ ("folderextension.crx", "folderextension.zip"), "2 Unit 42 file name", "3 oskeysetup.com lookup or download"), Detail = FolderPath),
    (DeviceNetworkEvents
    | where Timestamp > ago(30d) and RemoteUrl contains "oskeysetup.com"
    | extend Hit = "3 oskeysetup.com lookup or download", Detail = RemoteUrl)
| summarize Events = count(), FirstSeen = min(Timestamp), LastSeen = max(Timestamp), Details = make_set(Detail, 20) by DeviceName, Hit
| order by Hit asc

Hunt 5: cover tracks after a new MFA method


Unit 42 lists OAuth apps and workstation objects as persistence, and mailbox-rule changes as discovery and defense evasion. This flags any user with a registration and a new inbox rule, app consent or device registration on the same UTC day, our approximation of 24 hours.


Field mapping: inbox rules come from the Unified Audit Log, consent and device events from Entra.


False positives: day-one setup of new users.


Splunk


earliest=-30d latest=now ((index=YOUR_ENTRA_INDEX sourcetype="azure:monitor:aad" category=AuditLogs operationName IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info", "Consent to application", "Add device", "Register device")) OR (index=YOUR_O365_INDEX sourcetype="o365:management:activity" Operation IN ("New-InboxRule", "Set-InboxRule", "UpdateInboxRules")))
| eval what=coalesce(operationName, Operation)
| eval kind=if(in(what, "User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"), "reg", "follow")
| eval user=lower(if(kind="reg", coalesce(mvindex('properties.targetResources{}.userPrincipalName', 0), 'properties.initiatedBy.user.userPrincipalName'), coalesce('properties.initiatedBy.user.userPrincipalName', UserId)))
| eval day=floor(_time/86400)*86400
| stats count(eval(kind="reg")) AS registrations values(eval(if(kind="follow", what, null()))) AS follow_ons BY day user
| where registrations>0 AND mvcount(follow_ons)>0
| sort - day

Elastic


FROM logs-azure.auditlogs-*, logs-o365.audit-*
| WHERE @timestamp >= NOW() - 30 days
| EVAL what = COALESCE(azure.auditlogs.operation_name, o365.audit.Operation)
| WHERE what IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info", "Consent to application", "Add device", "Register device", "New-InboxRule", "Set-InboxRule", "UpdateInboxRules")
| EVAL reg = CASE(what IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"), 1, 0), follow = CASE(what IN ("User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"), NULL, what)
| EVAL user = TO_LOWER(COALESCE(CASE(reg == 1, `azure.auditlogs.properties.target_resources.0.user_principal_name`, NULL), azure.auditlogs.properties.initiated_by.user.userPrincipalName, o365.audit.UserId))
| EVAL day = DATE_TRUNC(1 day, @timestamp)
| STATS registrations = SUM(reg), follow_ons = VALUES(follow) BY day, user
| WHERE registrations > 0 AND follow_ons IS NOT NULL
| SORT day DESC

CrowdStrike


// needs Entra ID audit logs and the M365 Unified Audit Log ingested into LogScale or Next-Gen SIEM; #repo and every field name here are placeholders for your parser
// set the time picker to the last 30 days
#repo=YOUR_M365_REPO
| case {
    in(field=operation, values=["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]) | reg := 1 ;
    in(field=operation, values=["Consent to application", "Add device", "Register device", "New-InboxRule", "Set-InboxRule", "UpdateInboxRules"]) | reg := 0 | follow := operation }
| day := formatTime("%Y-%m-%d", field=@timestamp, timezone="UTC")
| groupBy([user_principal_name, day], function=[sum(reg, as=registrations), count(follow, as=follow_count), collect([follow])])
| test(registrations > 0)
| test(follow_count > 0)
| sort(day, order=desc)

Microsoft KQL


let regops = dynamic(["User registered security info", "User registered all required security info", "Admin registered security info", "User changed default security info", "User updated security info", "Admin updated security info"]);
union
    (AuditLogs
    | where TimeGenerated > ago(30d) and (OperationName in (regops) or OperationName in ("Consent to application", "Add device", "Register device"))
    | extend User = tolower(iff(OperationName in (regops), coalesce(tostring(TargetResources[0].userPrincipalName), tostring(InitiatedBy.user.userPrincipalName)), coalesce(tostring(InitiatedBy.user.userPrincipalName), tostring(TargetResources[0].userPrincipalName)))), Kind = iff(OperationName in (regops), "reg", "follow"), What = OperationName),
    (OfficeActivity
    | where TimeGenerated > ago(30d) and Operation in ("New-InboxRule", "Set-InboxRule", "UpdateInboxRules")
    | extend User = tolower(UserId), Kind = "follow", What = Operation)
| summarize Registrations = countif(Kind == "reg"), FollowOns = make_set_if(What, Kind == "follow", 20) by User, Day = bin(TimeGenerated, 1d)
| where Registrations > 0 and array_length(FollowOns) > 0
| order by Day desc

Two checks have no query. Teams message content rarely shows in audit logs, so ask users to report extortion messages. Unit 42 lists Rclone among exfiltration tools; look for it on machines that don't normally run it, knowing some IT teams use it for backups.


Indicators from Unit 42


Type

Indicator

Unit 42 context

SHA-256

930898e8b65e9cf4e2595ac4092042a707cca8922751e723ad8b86260655154a

folderextension.crx, the fake "Google Chrome" extension

SHA-256

ea8d30c42e758559e78bd1fe97f8849011beb7f24c7f092a066a5a62fa7fbe6f

folderextension.zip, same extension

URL

hxxps[:]//oskeysetup[.]com/files/folderextension.crx

Hosts the .crx above

URL

hxxps[:]//oskeysetup[.]com/files/folderextension.zip

Hosts the .zip above

IPv4

31.42.184[.]213

Hosted oskeysetup[.]com

IPv4

193.149.176[.]62

Accessed compromised accounts

User-Agent

python-requests/2.34.2

Scripted SharePoint and OneDrive collection (Cinder)

User-Agent

kiota-dotnet/1.17.1

Listed by Unit 42

User-Agent

Microsoft.Graph.Client/6.2.0

Listed by Unit 42

Kit endpoint

api_FyekIDWY.php

Seen on multiple domains in Com-affiliated cases

FQDN (1 to 7 of 56)

add-mfa[.]com, login-options[.]com, apply-passkey[.]com, auth-sso[.]com, authenroll[.]com, configkeys[.]com, configurekey[.]com

Unit 42 IOC list

FQDN (8 to 14 of 56)

editmysso[.]com, enableoskey[.]com, enrollmfa[.]com, enrollmfamethods[.]com, enrollssopasskey[.]com, fastpasskeys[.]com, hostsso[.]com

Unit 42 IOC list

FQDN (15 to 21 of 56)

installpasskey[.]com, installsso[.]com, itkeyenroll[.]com, itkeyportal[.]com, itkeyupdate[.]com, keyconfiguration[.]com, keyenrollos[.]com

Unit 42 IOC list

FQDN (22 to 28 of 56)

keyosregister[.]com, mfa-options[.]com, mfa-passkey[.]com, mfa-registry[.]com, mfaoptions[.]com, mfapasskeysetup[.]com, mfaregister[.]com

Unit 42 IOC list

FQDN (29 to 35 of 56)

mfasettings[.]com, my-passkey[.]com, my-passkeys[.]com, myaccountapps[.]com, myaccountoptions[.]com, myapps2fa[.]com, mypasskeyapp[.]com

Unit 42 IOC list

FQDN (36 to 42 of 56)

myssoapps[.]com, myssoauth[.]com, myssosettings[.]com, new-passkey[.]com, nowsso[.]com, onboardpasskey[.]com, oskeysetup[.]com

Unit 42 IOC list

FQDN (43 to 49 of 56)

oursso[.]com, passkey-mfa[.]com, passkeyconnect[.]com, register-mfa[.]com, register-passkeys[.]com, registermymfa[.]com, registersso[.]com

Unit 42 IOC list

FQDN (50 to 56 of 56)

settingsportal[.]com, signinoptions[.]com, ssokeyportal[.]com, ssosettings[.]com, startmypasskey[.]com, configuresso[.]com, makepasskey[.]com

Unit 42 IOC list


If you find one


Check a Test 1 row with the user first; a Hunt 1, 2 or 5 hit on that account, or an unexplained Hunt 3 or 4 hit, means acting now. Revoke the user's sessions and refresh tokens, remove the attacker's MFA method, review app consents and mailbox rules, and preserve the Unified Audit Log before it ages out. Unit 42 has seen these actors return weeks after a failed attempt.


One next step


For a second look at how your tenant handles MFA enrollment and sign-in risk, start with a free Inception Foresight M365 Assessment.


bg-map-white.png

INCEPTION SECURITY™

A cybersecurity company with in depth knowledge of the threat landscape and security controls.

NAVIGATION

GET IN TOUCH

© 2025 All Rights Reserved by INCEPTION SECURITY™ .

bottom of page