What is Vishing?
Vishing, short for voice phishing, is social engineering carried out over a phone call or another voice channel. The caller poses as IT support, a bank or a colleague and pushes the target to read out a code, reset a credential, approve an MFA prompt or install a remote access tool. MITRE ATT&CK files it as Spearphishing Voice, T1566.004.
Definition
- Vishing
- Vishing, short for voice phishing, is social engineering carried out over a phone call or another voice channel. The caller poses as IT support, a bank or a colleague and pushes the target to read out a code, reset a credential, approve an MFA prompt or install a remote access tool. MITRE ATT&CK files it as Spearphishing Voice, T1566.004.
How Vishing Works
The call happens outside your logs
In vishing the lure is a person talking, and phone systems rarely feed a SIEM. What you get is the consequence: a password reset, a new MFA method, a remote support tool that was not there yesterday. MITRE ATT&CK describes the technique as Spearphishing Voice (T1566.004), manipulating a user into providing access through a phone call or other voice communication, usually while posing as a trusted source and creating urgency.
The same page describes a callback variant. The victim receives a message that tells them to call a number, and on the phone they are walked to a malicious URL, a download, or the installation of a remote management tool. ATT&CK also notes that adversaries pair voice phishing with MFA request generation (T1621), so the caller keeps talking while the user accepts a push prompt the caller is triggering. This is social engineering with a voice attached, and it is the reason a plain phishing filter never sees it.
Three scripts, three different first traces
Callers vary the story, so classify each script by the system that records its first consequence.
| Script | What the caller wants | First trace you can query |
|---|---|---|
| "IT support" rings the employee | The employee installs or starts a remote support tool | Process creation of that tool on the endpoint (DeviceProcessEvents), the pattern ATT&CK records for Quick Assist (S1209) |
| "An employee" rings the help desk | The agent resets a password or registers a new MFA method | Entra ID audit events Reset password (by admin) and Admin registered security info; event 4724 on domain controllers |
| "IT support" rings and starts MFA prompts | The employee approves a push to make the noise stop | A burst of MFA prompts, then one success, in the sign-in log (T1621) |
| "The bank" or "a vendor" rings finance | Bank details, a payment, a one-time code read aloud | Usually nothing in telemetry. The evidence is an accounts payable ticket or a mailbox thread |
In the second row the attacker never touches your network. A help desk agent with legitimate rights makes the change, and the audit log faithfully records an authorised admin doing their job.
Two cases ATT&CK documents
Storm-1811 is tracked by ATT&CK as a financially motivated entity linked to Black Basta ransomware. It overloads a victim's inbox with non-malicious spam, then calls to offer help. ATT&CK records that it used Quick Assist during these calls to interact with victims and install follow-on malicious software.
Scattered Spider, in campaign C0027, impersonated legitimate IT personnel in phone calls to steer victims toward downloading a remote monitoring and management (RMM) tool. The same campaign registered devices for MFA to keep access through the victims' VPN, which is the footprint of the help desk row above.
A simulated sequence
Simulated example generated by SOCSimulator Research. The names, times and addresses are invented.
| Time | Source | What it shows |
|---|---|---|
| 09:41 | Help desk ticket HD-48211 | Phone call, "Dana Okoye is locked out before a client meeting" |
| 09:47 | Entra ID AuditLogs | Reset password (by admin), actor agent.lopez@example.com, target d.okoye@example.com |
| 09:52 | Entra ID SigninLogs | Successful sign-in for d.okoye@example.com from 203.0.113.58, an ASN the account has never used |
| 09:58 | Entra ID AuditLogs | Admin registered security info for the same target, a new phone number |
No single line here is an alert. Read together they show a reset, a sign-in five minutes later from somewhere new, and a new MFA method. The real Dana Okoye was never locked out.
Query: a reset or new MFA method followed by a sign-in
Run this in Microsoft Sentinel. It pairs each administrator-driven reset or security-info change with a successful sign-in by the same account within two hours.
let lookback = 1d;
let followWindow = 2h;
AuditLogs
| where TimeGenerated > ago(lookback)
| where OperationName in ("Reset password (by admin)", "Admin registered security info")
| extend Target = tostring(TargetResources[0].userPrincipalName),
Actor = tostring(InitiatedBy.user.userPrincipalName)
| project ChangeTime = TimeGenerated, OperationName, Actor, Target
| join kind=inner (
SigninLogs
| where TimeGenerated > ago(lookback)
| where ResultType == 0
| project SignInTime = TimeGenerated, UserPrincipalName, IPAddress, AppDisplayName, Location
) on $left.Target == $right.UserPrincipalName
| where SignInTime between (ChangeTime .. (ChangeTime + followWindow))
| project ChangeTime, SignInTime, OperationName, Actor, Target, IPAddress, Location, AppDisplayName
| order by ChangeTime descMicrosoft lists both operation names in its Entra audit activity reference, and the on-premises counterpart is event 4724, generated when an account attempts to reset the password of another account. SOCSimulator has not run the query against a live tenant. Most of its rows will be routine resets, so the work is the filter you add next: compare IPAddress with the user's previous 30 days and keep only the sign-ins from a place the account has not been.
Query: a remote support tool that is new on the device
Run this in Defender XDR advanced hunting. The image names are a starting list, not a complete one, and you should replace them with whatever your estate does and does not allow.
let tools = dynamic(["quickassist.exe", "anydesk.exe", "teamviewer.exe"]);
let seenBefore = DeviceProcessEvents
| where Timestamp between (ago(30d) .. ago(1d))
| where FileName in~ (tools)
| extend FileName = tolower(FileName)
| distinct DeviceName, FileName;
DeviceProcessEvents
| where Timestamp > ago(1d)
| where FileName in~ (tools)
| extend FileName = tolower(FileName)
| join kind=leftanti seenBefore on DeviceName, FileName
| project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine
| order by Timestamp descAdvanced hunting keeps 30 days of data, so a tool that last ran two months ago looks new. The first run of a remote tool on a laptop that belongs to someone with no support role is the part worth a phone call.
Vishing next to its neighbours
Smishing arrives by text and phishing by email, while vishing comes by voice. The three chain: ATT&CK's own description has a message telling the victim to call a number, which turns a text into a vishing call. The controls differ too. Email gateways and URL filters stop the first two. Nothing in the mail path stops a caller, so the control that works is procedural: the help desk confirms identity through a channel the caller did not choose.
Vishing in SOC Operations
Start from the change, not from the sign-in. When a password reset or a new MFA method is followed by a login from a place the account has never used, find the ticket or call record behind the reset and ask the help desk who called, which number they rang back and what proof the caller gave. A callback to a number the caller supplied means the reset was never verified. Then work outward: revoke sessions, remove the newly registered method, reset the password through a channel you control, and review mailbox rules and OAuth consents on the account. If the call went to an end user, list the remote tools that started on their host in the hour after it, and ask what they were told to type or click, because the script is evidence. Record the pretext, the number if known, the minutes between call and first sign-in, and the tool used. That gives the help desk a control to tighten and gives hunters a lead for the other staff who were called that morning.
Practice Vishing in a Real SOC
SOCSimulator provides hands-on training with realistic SIEM, XDR, and Firewall interfaces. Build real analyst skills investigating vishing scenarios with zero consequences, free.
Frequently Asked Questions
- How does a vishing attack work?
- The attacker calls with a pretext, such as IT support fixing a problem or an employee locked out of an account, and creates urgency. The goal is a concrete action: installing a remote support tool, accepting an MFA prompt, reading out a code, or getting a help desk agent to reset a password or register a new MFA method. The call itself leaves no log in the SIEM. The reset, the new method or the tool does.
- What is the difference between vishing and phishing?
- Phishing reaches the target through written messages, mostly email, so gateways and URL filters get a chance to stop it. Vishing is a live conversation, which lets the attacker answer objections and adapt. ATT&CK treats voice phishing as a variant of spearphishing, and the two chain: a message that tells the victim to call a number is phishing that ends in a vishing call.
- What is the difference between vishing and smishing?
- The channel. Smishing is a lure sent by SMS or another text message, and vishing is one delivered by voice. Smishing leaves almost nothing in corporate logs until the sign-in that follows, and vishing leaves even less before the first consequence. A text that says to call a number to dispute a charge starts as smishing and finishes as vishing.
- What is an example of vishing?
- In campaign C0027, ATT&CK records Scattered Spider impersonating legitimate IT personnel in phone calls to direct victims to download a remote monitoring and management tool. Storm-1811 used a similar script: it flooded a mailbox with spam, then called offering help and used Quick Assist to interact with the victim and install follow-on software.
- How can you prevent vishing?
- Take identity out of the caller's hands. The help desk verifies through a channel it already holds, such as a call back to the directory number, before any password or MFA change. Staff do not install remote tools because a caller asks, and they have a short route to report the call. ATT&CK's mitigation for this technique is training users to be suspicious of callers and to verify who they are.
Sources
- MITRE ATT&CK: Phishing: Spearphishing Voice (T1566.004)
- MITRE ATT&CK: Storm-1811 (G1046)
- MITRE ATT&CK: Scattered Spider campaign C0027
- Microsoft Learn: Microsoft Entra audit log activities reference
- Microsoft Learn: Event 4724, an attempt was made to reset an account's password
- Microsoft Learn: Advanced hunting overview (Defender XDR)
Related Terms
In cyber security, social engineering means manipulating a person into handing an attacker access, i...
Phishing is a social engineering attack delivered via email, SMS, voice calls, or other channels tha...
MFA (multi-factor authentication) is a sign-in check that asks for two or more different kinds of pr...
An insider threat is a security risk originating from current or former employees, contractors, or b...
Incident response (IR) is how a security team detects, contains, removes and recovers from an incide...
EDR (endpoint detection and response) is security software that puts an agent on each laptop, server...