Business Email Compromise (BEC): How It Works and Detection
Skip to content

Business Email Compromise: Stages, Detection and Response

Business email compromise (BEC) is fraud carried out through a trusted mailbox. An attacker takes over or imitates a business email account, then uses it to redirect a payment, change bank details, or collect sensitive information.

The fraud is the last step. Before it, the attacker signs in, hides, and reads. This guide follows those stages in order and shows what each one leaves behind in Microsoft 365, and how to detect it.

How Business Email Compromise Works

A BEC attack succeeds because the request looks genuine. It arrives in a real conversation, from a real address or a near-identical one, at the moment a payment is expected.

In some cases the attacker controls a real mailbox, inside the organization or at one of its suppliers. In others the attacker controls nothing and relies on imitation: a lookalike domain, or a display name that matches an executive.

TypeHow it worksWhere the evidence is
Invoice fraudA real invoice thread is intercepted and resent with new bank details.A compromised mailbox on one side of the conversation.
CEO fraudA message that appears to come from an executive asks for an urgent payment.Often a lookalike domain or display name. No mailbox may be compromised.
Vendor email compromiseA supplier’s mailbox is taken over and used against its customers.The supplier’s tenant. The customer holds only the received messages.
Payroll diversionA request arrives to move an employee’s pay to a new bank account.Usually imitation, sometimes the employee’s real mailbox.
Data theftThe mailbox is mined for identity documents, contracts, or credentials.Mail access records in the organization’s own tenant.

The mailbox-takeover form is the more damaging, and the more detectable. The stages below describe it.

How Large the Problem Is

Business email compromise receives less attention than ransomware and costs more. It causes no outage and is often discovered only when a supplier asks why an invoice is unpaid.

The FBI’s Internet Crime Complaint Center recorded more than USD 3 billion in reported BEC losses from more than 24,000 complaints in its 2025 annual report, and USD 2.77 billion from more than 21,000 complaints the year before. Those figures cover only incidents reported to the FBI.

In Australia, the Australian Signals Directorate’s Annual Cyber Threat Report 2024-25 lists email compromise and BEC fraud among the most commonly reported cybercrimes for businesses. Email compromise with no financial loss made up 19 percent of self-reported business cybercrime, and BEC fraud with financial loss a further 15 percent.

BEC is attractive to criminals on payout, skill, and risk. The payout from a single diverted invoice can exceed a typical ransom. The technical skill required is low, since phishing kits and stolen sessions are sold as services. The risk is small, because the fraud looks like an ordinary business payment until someone checks.

Who Carries Out BEC

BEC is organized work, split across people with different skills.

Kit developers

They build and rent the phishing platforms that relay a sign-in and capture the session. Subscribers get hosting, page templates that copy the Microsoft sign-in, and a dashboard of captured sessions.

Phishers

They run the campaigns: buying or scraping target lists, sending the lures, and collecting working sessions. Many sell the access on and never touch the fraud.

Mailbox operators

They are patient readers who work through a compromised mailbox, learn the payment cycle, and prepare the rules and lookalike domains.

Fraud writers

They compose the message. Fluent, business-like writing in the victim’s own tone matters more than any technical step.

Money mules and launderers

They provide the receiving bank accounts and move the funds onward before a recall arrives.

The separation matters to an investigation. The person who phished the account and the person who sent the fraudulent invoice may be different, working days apart from different networks. A gap between the first suspicious sign-in and the next burst of activity often marks the point where access changed hands.

It also explains the speed at each end. Securing access takes minutes and is nearly automatic. The reading stage in the middle is slow and manual, which is where defenders have time.

Why Email Security Misses It

Email filters look for malicious attachments, known bad links, and senders that fail authentication. A BEC message usually has none of these.

It comes from a real mailbox

When a supplier’s account is compromised, the message is sent from the genuine address, through the genuine mail system, and passes every sender check.

It contains no malware

The payload is a sentence asking for a payment to a new account.

It sits inside a real thread

The conversation history is authentic, and the reply continues it.

The link, where there is one, leads to a legitimate service

Phishing pages are often hosted on, or reached through, well-known file-sharing and form platforms.

Internal messages skip the filter

A message from a compromised colleague may never pass through external mail scanning at all.

This is why detection has to move from the message to the account. The sign-in and audit records described in the stages below show the compromise even when the email itself looks perfect.

The Stages of a BEC Attack

StageWhat the attacker is doingBest source of visibility
1. TargetingChoosing people who request, approve, or receive payments.Email security logs, lookalike domain monitoring.
2. Account accessSigning in with a stolen password or a stolen session.Entra ID sign-in logs.
3. PersistenceMaking sure a password reset does not remove access.Entra ID audit log, and the Microsoft 365 audit log for delegation and forwarding.
4. Mailbox reconnaissanceReading and searching mail to learn how money moves.Microsoft 365 audit log, mailbox audit records.
5. ConcealmentHiding replies and warnings from the real user.Inbox rule and mailbox setting changes in the audit log.
6. FraudSending the payment request or changed bank details.Message trace, the finance process itself.
7. Cash-out and reuseMoving the money and phishing the victim’s contacts.Bank records, outbound mail volume.

Stages 2 to 5 can last weeks. Every one of them is recorded in the tenant, if the logs are kept and someone is looking.

Stage 1: Targeting

Attackers pick targets by role. Accounts payable staff, payroll, executive assistants, finance managers, and salespeople who send invoices are the most useful mailboxes. Names and roles come from professional networking sites and the organization’s own website.

The first contact is usually a phishing email built around something the target expects: a shared document, a voicemail notification, a payment remittance, or a request to re-authenticate. Increasingly it comes from another compromised organization the target already deals with.

Detection methods:

Sender authentication results

Messages that fail or only partly pass the sender’s published authentication policy are recorded in the message headers and the email security log.

First-time senders in a payment context

A domain that has never written to the organization before sends an invoice or a document link to finance.

Newly registered domains

The sender or link domain was registered within the past days or weeks.

Lookalike domains

A registration differs from the organization’s domain, or a major supplier’s, by one character, a swapped pair of letters, or a different ending.

Links to sign-in pages

The link leads, through redirects, to a page that asks for Microsoft 365 credentials on a domain that is not Microsoft’s.

User reports

A reported phishing message is the earliest warning available. Check who else received it and who clicked.

Stage 2: Account Access

Many mailbox takeovers now get past multi-factor authentication by stealing the session that follows it, through a phishing page that relays the sign-in. The session hijacking guide explains the mechanism step by step.

Other routes still work. Passwords reused from another breach are tried against accounts without a second factor. Users are tricked into entering a device sign-in code, or into granting a malicious application access to their mailbox. Basic authentication is switched off in Exchange Online for every protocol except SMTP AUTH, so password-only mail protocols remain a route mainly on on-premises Exchange and non-Microsoft mail systems.

Detection methods:

The first sign-in comes from the wrong place

With a relay phishing page, the sign-in that the user completed appears in the log from the attacker’s server, usually a hosting provider or anonymizing network.

One session, two networks

The same session appears minutes later from a second address, with a different browser and no new authentication.

Identity protection detections

Microsoft Entra ID Protection raises detections such as unfamiliar sign-in properties, anonymous IP address, impossible travel, anomalous token, and attacker in the middle.

Password-only protocol sign-ins

A successful SMTP AUTH sign-in, or any basic authentication sign-in to an on-premises or non-Microsoft mail system, has skipped the second factor.

The account takeover guide shows how to read the device, client, and authentication fields of a sign-in record, and why the non-interactive log holds most of the attacker’s continued use.

Stage 3: Persistence

A stolen session expires, and a suspicious user changes their password. Attackers act within minutes of the first sign-in to make their access survive both.

The most common step is registering their own authentication method on the account: an authenticator app or a phone number. From then on, the attacker can complete multi-factor authentication independently. Other methods include registering a device, granting a third-party application ongoing access to the mailbox, delegating the mailbox to another account, and adding a forwarding address.

Detection methods:

New authentication method after a risky sign-in

The Entra ID audit log records security information being registered. A new method added minutes after a sign-in from an unfamiliar network means an attacker is securing access.

Application consent

A user consents to an application that requests permission to read mail, particularly with offline access, from a publisher the organization does not know.

Device registration

A new device is registered to a user from an unfamiliar address.

Mailbox delegation and forwarding

The Microsoft 365 audit log shows another account granted full access or send-as rights, or a forwarding address set on the mailbox.

Password change by the attacker

A self-service password change from the attacker’s address locks the real user out and is sometimes the first thing anyone notices.

A password reset removes none of these. This is why accounts are compromised again days after an incident was thought to be closed.

Stage 4: Mailbox Reconnaissance

The attacker now studies the business. They search the mailbox for words such as invoice, payment, remittance, wire, and bank. They read the threads those searches return and learn who the suppliers are, who approves payments, what the invoices look like, and when the next large payment is due.

Some attackers read through the web mail interface. Others connect a mail client or a synchronization tool and download the entire mailbox, which gives them every message and attachment to work through offline.

Detection methods:

Mail access records

The Microsoft 365 audit log records mailbox access as MailItemsAccessed events. Access by an attacker’s session shows which messages and folders were opened.

Full synchronization

A sync-type access record from an unfamiliar client means the mailbox was downloaded. Sync is recorded per folder, so every item in a synced folder is treated as copied.

Mailbox searches

Search activity for payment terms from an unusual session is recorded where search auditing is available.

Access volume

The mailbox is opened far more heavily than the user’s normal pattern, or at hours when the user is not working.

Access to shared mailboxes

The user’s session opens the accounts payable or payroll shared mailbox for the first time.

Mail access records are the evidence that later answers the question every incident raises: what did the attacker read? A slow compromise can outlast the audit log’s default retention unless logs are sent to a SIEM or archive.

Stage 5: Concealment

The fraud only works if the real mailbox owner does not see the replies. Attackers create inbox rules to intercept them. The rules themselves are distinctive, and they are the most reliable single indicator of BEC.

What the rule doesWhy the attacker wants itWhat to look for
Moves messages from one sender or domain to an obscure folderReplies from the targeted supplier or customer never reach the inbox.Rules that move mail to folders such as RSS Feeds, Conversation History, or Archive.
Moves or deletes messages containing certain wordsHides anything about the payment, and any warning from IT or the bank.Conditions on words such as invoice, payment, hacked, phishing, or password.
Marks matching messages as readStops unread counts drawing attention.Mark-as-read combined with a move.
Forwards mail to an external addressKeeps a copy of everything after access is lost.Forwarding or redirect actions to consumer mail domains.
Deletes everythingBuys time while a mass phishing run goes out from the account.A rule with no conditions that deletes all incoming mail.

Attackers often give these rules a name made of one or two punctuation marks or a single letter, so they are easy to miss in a list. A rule with a name like that deserves immediate review.

Detection methods:

Rule creation and changes

The audit log records new and modified inbox rules. Review every rule that moves, deletes, forwards, or marks as read based on a sender or keyword.

Rules created from an unfamiliar address

The client address on the rule creation matches the attacker’s sign-in and differs from the user’s usual networks.

Mailbox-level forwarding

A forwarding address is set on the mailbox itself, where it does not appear in the user’s rule list.

Deleted sent items

Messages are sent and then removed from Sent Items and Deleted Items within minutes.

Tenant-wide review

One phishing campaign usually compromises several people, so check every mailbox.

Stage 6: The Fraud

With the groundwork done, the attacker waits for a real transaction and steps into it. They reply inside the existing thread, with the genuine invoice attached and one change: the bank account.

Many attackers move the conversation to a lookalike domain at this point. The reply-to address differs from the genuine one by a character, so the rest of the exchange happens outside the compromised mailbox and continues even after it is secured.

Detection methods:

Reply-to mismatch

The reply address differs from the sender address, or from the address used earlier in the thread.

Mid-thread domain change

A participant’s domain changes partway through a conversation.

Unusual sending pattern

Messages go out from the account at odd hours, from the attacker’s address, to finance contacts.

Internal requests

A message from a compromised executive’s real account to finance carries no external warning banner, so staff need a process they follow regardless of sender.

The control that stops the payment at this stage belongs to the finance team, and the section after Stage 7 describes it.

Stage 7: Cash-Out and Reuse

Once funds arrive, they are moved onward quickly, often through several accounts and across borders.

The mailbox is then reused. Attackers send phishing from it to the victim’s whole address book. Recipients trust the sender, and the cycle begins again in their organizations. This mass send is frequently how a compromise is finally noticed.

Detection methods:

Outbound volume

The account sends hundreds of messages in a short period, or hits sending limits.

Identical messages to many recipients

The same link or attachment goes to external contacts.

Bounce and complaint spikes

Non-delivery reports rise suddenly, along with replies asking whether a message is genuine.

What the Finance Team Sees

The attack looks different from the accounts payable desk, and the people there are the last line of defense.

A typical case runs as follows. A supplier the company has paid for years sends an invoice for a real order. Two days later a reply arrives in the same thread. It apologizes for the confusion, explains that the supplier’s bank account is under audit, and attaches the same invoice with new account details. It asks for confirmation once the payment is made.

Everything matches. The sender’s name, the signature block, the invoice number, the amount, and the writing style are all correct. The only change is the account, and often the bank is in the same country.

Details that finance staff can check, without any technical tools:

The reply address: Compare it with the address used in earlier messages, character by character.

The reason for the change: Audits, tax issues, and bank migrations are the standard stories.

The account name and location: Check that the new account is in the supplier’s name and usual country.

The pressure: Note whether the message asks for speed, a shortcut around the usual process, or secrecy.

The channel: Ask the supplier about the change on a phone number taken from the contract or a previous invoice.

The call-back is the control that works, backed by a second approver for new or changed payees. It has to use a number the organization already holds, since the number in the email belongs to the attacker.

Visibility You Need

BEC is detected almost entirely from cloud logs. No endpoint agent sees it, because nothing runs on the user’s device.

SourceWhat it recordsStages coveredDefault retention
Entra ID sign-in logsInteractive and non-interactive sign-ins, with address, device, client, and authentication detail.2, 430 days with paid licenses, 7 days on the free tier.
Entra ID audit logAuthentication method changes, application consent, device registration, password changes.330 days with paid licenses, 7 days on the free tier.
Microsoft 365 audit logMail access, inbox rules, forwarding, delegation, sent and deleted items.3, 4, 5, 6180 days.
Message trace and email security logsDelivery, sender authentication results, and messages sent from the account.1, 6, 7Varies by product, often days to weeks.
Lookalike domain monitoringNew registrations that imitate the organization or its suppliers.1, 6External service.

Most failed investigations come down to retention or review. Sign-in logs expire before anyone looks, so the first sign-in can no longer be found. Non-interactive sign-ins and mailbox audit records go unreviewed because no alert points to them.

Sending these logs to a SIEM solves both. It extends retention and lets one detection connect a risky sign-in, a new authentication method, and an inbox rule on the same account within the same hour.

Detection Priorities

The detections are ranked here by how early they fire and how rarely they are wrong. The account takeover guide covers how to build each one as a sequence.

1
Risky sign-in, then a new authentication methodThis fires within minutes of the takeover, before any mail is read.
2
Inbox rules that hide or forward mailThis produces the fewest false alarms of any detection on the list.
3
One session used from two networksThis confirms a stolen session without waiting for the attacker’s next move.
4
Application consent for mailbox accessThis catches the foothold that a password reset leaves behind.
5
Mailbox synchronization from an unfamiliar clientThis marks the point where the whole mailbox counts as copied.
6
External forwardingThis is rare in most tenants and cheap to review every time.
7
Outbound mail spikesThis fires last, when the compromise is already being reused.
8
Sign-ins from hosting and anonymizing networksThis is the noisiest on the list and works best as a condition on the others.

When a Detection Fires

1
Block sign-in if the attacker is active or the account is privilegedKeep it blocked until the review is done.
2
Reset the password, then revoke sessionsThe reset comes first so the stolen password is useless by the time the sessions end.
3
Remove what the attacker addedRecord each item, then remove it: authentication methods, inbox rules, forwarding, delegation, application grants, and registered devices.
4
Call the bankIf a payment has gone, ask the bank’s fraud team for a recall and a freeze on the receiving account immediately. Hours matter, so this call runs in parallel with the account steps.
5
Hold related paymentsTell finance to stop payments connected to the mailbox, and review recent changes of bank details.
6
Scope the tenantSearch every account for sign-ins from the attacker’s addresses and for similar inbox rules.
7
Establish what was read and sentUse mail access records and message trace across the whole period of access.
8
Warn contactsTell suppliers and customers who received mail from the account, by phone where a payment is in progress.
9
Confirm in the sign-in logsCheck that nothing from the attacker’s sources succeeds afterward.

The account takeover guide explains what each account step removes.

Report the fraud to police as well as the bank. In the United States, file a complaint with the FBI’s Internet Crime Complaint Center at ic3.gov as soon as the fraud is found. Its Recovery Asset Team works with banks to freeze funds, and the FBI’s process for international wires depends on the transfer being reported within 72 hours. In Australia, cybercrime reports are made through ReportCyber at cyber.gov.au.

Where the mailbox held personal information and the attacker accessed it, a data breach assessment is needed. In the United States, state breach notification laws decide who must be told. In Australia, an eligible data breach under the Privacy Act must be notified to the Office of the Australian Information Commissioner and the affected individuals.

Controls That Break the Chain

The account takeover guide maps each step of a takeover to the control that breaks it, including phishing-resistant authentication and device requirements. The controls specific to mailbox fraud are these.

External forwarding blocked

Automatic forwarding to outside domains is set to Off in the outbound spam policy. The system-controlled default behaves differently between tenants, so the setting is made explicit.

Alerts on mailbox changes

New inbox rules, forwarding addresses, and delegation raise an alert for review.

Application consent restricted

Users cannot grant mailbox access to third-party applications without review.

Password-only mail sign-in removed

SMTP AUTH is disabled for mailboxes that do not need it, and on-premises or non-Microsoft mail systems require modern authentication.

Sender authentication enforced

The organization’s own domains publish a DMARC policy of reject, so the exact domain cannot be spoofed to customers and suppliers.

External sender marking

Mail from outside the organization is tagged, which makes a lookalike domain posing as a colleague easier to spot.

Supplier contact records

A verified phone contact is recorded for each supplier at onboarding, so the call-back number exists before it is needed.

Payment verification

Every change of bank details is confirmed by call-back, and new or changed payees need a second approver.

BEC Detection and Response With ThreatDefence

ThreatDefence monitors Microsoft 365 mailboxes and Entra ID sign-ins 24x7. Our identity detections follow the stages in this guide: sign-ins that break an account’s pattern, new inbox rules and forwarding, application consents, and changes to authentication methods. AI agents check each alert against the account’s history, and our SOC disables the account and revokes its sessions under agreed authority.

When fraud has already happened, our responders establish when access began, which messages were read, and which rules and grants must be removed, so finance and the bank receive a dated account of events. A free Microsoft 365 forensic review examines six months of tenant activity for compromise that went unnoticed.

A Microsoft 365 forensic scan in the ThreatDefence portal: six months of activity analyzed, sign-ins by risk, MFA coverage, app consents, findings by severity and the response taken

Experiencing a breach?

Our incident response team answers 24x7, with or without a retainer.

Talk to Our Experts

Ask about a free Microsoft 365 forensic review, covering six months of activity in your tenant.

    I'm an MSP, MSSP, or reseller

    By submitting this form, you accept the terms in our privacy policy.

    Frequently Asked Questions

    How does a business email compromise attack work?

    An attacker gains access to a business mailbox, usually through phishing, then registers a sign-in method of their own, reads mail to learn how payments are made, hides replies with inbox rules, and sends changed bank details at the moment a real invoice is due. The preparation can run for weeks, and the fraud takes one email. A second form skips the mailbox and imitates the sender from a lookalike domain.

    How can business email compromise be detected?

    Detection comes from the account, since the fraudulent email passes every filter. The earliest signals are a sign-in from an unfamiliar network followed within minutes by a new authentication method, and an inbox rule that moves or forwards mail based on a sender or keyword. Both are recorded in Entra ID and Microsoft 365 audit logs days or weeks before any payment request is sent.

    What is the difference between BEC and phishing?

    Phishing is a delivery method: a deceptive message that steals credentials or a session, or delivers malware. Business email compromise is the fraud, where a trusted mailbox or identity is used to redirect money or obtain information. Phishing is usually how the mailbox is taken, and the taken mailbox is then used to phish the victim’s contacts, so the two alternate along a supply chain.

    Is CEO fraud the same as business email compromise?

    CEO fraud is one type of business email compromise. The attacker poses as a senior executive and pressures staff to make an urgent, confidential payment. It usually runs from a lookalike domain or a matching display name, with no mailbox in the organization accessed, so the sign-in logs hold nothing to find. The defense is a payment process that applies to requests from executives as well.

    Can business email compromise happen with MFA enabled?

    Yes. A phishing page that relays the sign-in captures the session issued after the second factor, and the attacker then registers a method of their own. Device code and application consent requests give access with no fake page at all. Impersonation from a lookalike domain needs no sign-in. Passkeys and security keys stop the relay, and only checking the address stops the lookalike domain.

    Which inbox rules indicate a compromised mailbox?

    Look for rules that move mail from one sender or domain to a folder such as RSS Feeds or Conversation History, delete or hide messages containing payment or security words, mark matching mail as read, or forward to an external address. A rule named with a single character or punctuation mark is a common attacker habit. A rule created from an address the user has never signed in from settles the question.

    How do we know which emails the attacker read?

    The Microsoft 365 audit log records mail access as MailItemsAccessed events, each tied to a session and a client address. Individual access records name the messages opened. Sync records mean a desktop mail client downloaded the folder, and the attacker should be assumed to hold a full copy of it. Where the logs have expired, the report says so and treats the period as possible access.

    How quickly must we act to recover a fraudulent payment?

    The same day, and preferably within hours. Call the bank’s fraud team as soon as the fraud is suspected and ask for a recall and a freeze on the receiving account. In the United States, file a complaint at ic3.gov at the same time so the FBI can contact the receiving bank. Funds are moved onward quickly, and the FBI’s process for international wires depends on a report within 72 hours.

    What is vendor email compromise?

    Vendor email compromise is a BEC attack that starts in a supplier’s mailbox. The attacker reads real invoice threads and sends the supplier’s customers new bank details from the genuine address. The customer’s own tenant holds no sign-in evidence, only the received messages, so the supplier has to investigate its side. For the customer, a call to a number on file is the only check that works.