Leaving this course?

Lesson 1 of 9

What it is, and why it is not phishing

Business email compromise, usually shortened to BEC, is fraud that arrives as a correct-looking instruction rather than as an attack. The criminal either takes control of a real email account or imitates one closely enough to pass, and then asks for a payment the business was going to make anyway, to a different account.

The United States financial intelligence unit, FinCEN, describes email compromise fraud as schemes in which criminals compromise the email accounts of victims in order to send fraudulent payment instructions to financial institutions or other business associates. Its advisory sets out two mechanisms. Criminals may compromise an account directly, through an unauthorised electronic intrusion, and send from it. Or they may impersonate an account, by spoofing the address or by using one that closely resembles a known counterparty's, altered by adding, changing or deleting one or more characters.

That second mechanism is why technical defences struggle with this. There may be no malware to find, no attachment to scan and no link to open in a sandbox. There is a plausible request, from a name the recipient knows, about a transaction that is really happening.

The scale. In 2024 the FBI's Internet Crime Complaint Center recorded 21,442 BEC complaints with reported losses of 2,770,151,146 dollars. That was the second-largest loss of any category it tracks, behind investment fraud. FinCEN's advisory of July 2019 put possible losses affecting United States financial institutions and their customers at over 9 billion dollars since 2016.

Those are American figures from American reporting systems, and the FinCEN advisory is now some years old, so treat the sums as indicative rather than current. The pattern is another matter. The processes under attack, which are supplier payments, payroll and property completions, exist in every economy, and a business in Nairobi or Mumbai with an email account and a bank account is exposed in the way a business in Texas is.

Three distinctions are worth fixing in your mind before going further:

  • BEC is not phishing. Phishing harvests credentials from many people at once. BEC uses stolen credentials, or a convincing imitation, against one chosen business process.
  • BEC is not an attack on the bank. The payment is properly authenticated and properly executed. Nothing in the bank's systems misbehaves.
  • BEC is not, in the first instance, a technology failure. It is the failure of a business process that accepts payment instructions by email.

Terminology: the FATF standards say suspicious transaction report (STR), while the United States and the United Kingdom say suspicious activity report (SAR). They are the same filing, and this course uses whichever term its own sources use.

Lesson 2 of 9

The business processes criminals target

BEC is not aimed at companies so much as at processes. The target is any routine in which money moves on the strength of an emailed instruction.

Changing a supplier's bank details. A message appears to come from a known supplier: our bank has changed, please use these details for the next invoice. If the finance team edits the supplier record and pays the next genuine invoice, the money is gone and the real supplier is still owed. This is the single most productive version of the fraud, because the invoice itself is legitimate.

The executive instruction. A message that appears to come from the chief executive or finance director asks for an urgent payment, often framed as confidential, often while that person is known to be travelling. FinCEN's analysis found that approximately half of all BEC fraud targeting financial institutions was carried out through emails impersonating the chief executive or president.

Payroll diversion. An employee appears to ask for their salary account to be changed. One salary is a small loss; the same request across a payroll department is not.

Property completions. A buyer receives apparently genuine closing instructions and wires the deposit to the criminal. FinCEN explains why this sector suits the fraud so well: public information about transactions and counterparties is available, the parties habitually communicate by email, and there is commonly no strong process for verifying the identity and validity of instructions.

Which sectors it reaches. FinCEN's reporting gives the three most commonly targeted as manufacturing and construction, at 25% of reported BEC cases, commercial services at 18%, and real estate at 16%. Dozens of government bodies have been hit as well, through pension accounts, payroll accounts and contracted services.

The education figure is the one to remember, because it shows the shape of the risk rather than its frequency. In 2016 financial institutions reported to FinCEN over 160 incidents of BEC targeting educational institutions, with attempts to steal over 50 million dollars. The sector accounted for only around 2% of BEC incidents in that reporting, yet it held the largest concentration of high-value attempts.

Read down that list and a pattern appears. Every process named shares three features: a payment is expected, the amount is routine enough not to prompt a question, and the instruction can arrive in writing from outside. Hunt for those three in your own organisation.

Lesson 3 of 9

Why the money is usually gone, and the hours that decide

The hardest thing to accept about BEC is that the payment is valid. The payer gave the instruction, the bank followed it, and the funds arrived where they were sent.

European law makes the consequence of that unusually clear. Under Article 64 of the second Payment Services Directive, a payment transaction is considered to be authorised only if the payer has given consent to execute it. A payer deceived into giving the instruction has still given consent, so the transaction is authorised, and the Directive's remedy for unauthorised transactions does not reach it. This is the legal hinge of the whole subject, and it is why losses fall where they do. Consumer reimbursement regimes have been built in some countries to soften it for individuals and small businesses; this course's companion on scam reimbursement covers them.

So recovery is not a legal question but a race. FinCEN puts it plainly: such transactions are often irrevocable, which leaves institutions and customers unable to cancel the payment or recall the funds, so identifying fraudulent payment instructions before the payment is issued is essential.

Where speed is possible, it works. The FBI operates a Recovery Asset Team, established in 2018, which asks a receiving bank to freeze funds through what it calls the Financial Fraud Kill Chain. In 2024 that process ran on 3,020 complaints covering 848.4 million dollars of attempted theft. Domestically, 2,651 complaints led to 469.1 million dollars frozen; internationally, 369 complaints led to 92.5 million frozen. The reported success rate was 66%.

Two cases from the same report show what the hours are worth. In March 2024 a property buyer wired 956,342 dollars on spoofed instructions from their supposed estate agents, realised two days later, and the team froze the receiving account: 955,060 dollars was stopped and returned. In September 2024 a request came through the FBI's Singapore office about a 6,661,650 dollar transfer; the receiving bank blocked the account and 5,100,000 dollars was held, while the rest had been wired onward to accounts in Spain and China immediately upon deposit.

That last clause is the lesson. The criminal's own process is built for speed, so yours has to be too. What determines recovery is almost never cleverness after the event. It is how quickly the victim notices, how quickly they telephone, and whether anyone knows who to telephone.

Lesson 4 of 9

The controls that actually work

Because BEC attacks a process, the effective controls are changes to that process. The technical ones help, and they are not where the leverage is.

Verify out of band, on a number you already hold. Any instruction to pay a new account, or to change an existing one, is confirmed by telephone to a contact already on file, never to a number, address or extension given in the message itself. FinCEN's own recommendation is to verify suspicious emailed instructions by using multiple means of communication, or by contacting others who are authorised to conduct the transaction. One rule covers most of it: the channel used to verify must be chosen by you, not by the sender.

Treat a change of bank details as a new payee, not an edit. An edit inherits the trust built up by every previous successful payment. A new payee starts with none. Systems that log a bank-detail change as a new relationship, with its own approval, remove the fraud's main advantage.

Separate who may change a record from who may approve a payment. One person requests, a second approves, and the approver sees that the details changed recently. Dual authorisation on amendments to the supplier master file matters more than dual authorisation on payments.

Delay the first payment to a new account. A short hold on the first payment, or a low limit on it, converts an irreversible loss into a recoverable one. It costs a little goodwill and is the single most valuable control on this list.

Check the name against the account. Where a banking system offers a payee-name check before the payment leaves, use it, and act on a mismatch rather than clicking past it.

Train for the specific thing. General security awareness does not prepare anyone for a courteous message from a real supplier's real address about a real invoice. Training should use your own invoices, your own supplier names and your own approval screens, and it should state explicitly that urgency and confidentiality in a payment request are themselves warning signs.

Tell the other side. The United Kingdom's National Cyber Security Centre suggests telling suppliers and customers what to expect from you, in words as plain as: our bank details will not change at any point. It gives whoever receives a forged message in your name a second chance to notice.

Write down the one rule people need under pressure. Staff will not recall a policy at the moment it matters; they will recall a sentence. Something of the form: we never change payment details on the strength of an email, however convincing, without a call to the number we already hold.

Notice what is not on this list: no detection model and no email filter. Both are worth having. Neither saves a company whose process accepts instructions from whoever sent the last message.

Lesson 5 of 9

What a bank can see

A bank cannot read its customer's invoices or know which supplier is genuine. It can see the payment, and a BEC payment often looks different from the customer's ordinary behaviour.

On the sending side, the signals worth scoring are mostly about novelty and pace:

  • a first payment to a beneficiary never paid before, especially for a large or round amount;
  • a beneficiary account in a country the customer has never paid, or a domestic account where this supplier has always been paid abroad, or the reverse;
  • a payment that breaks the customer's rhythm, such as a monthly supplier paid twice in a week, or a payment outside the hours this customer works;
  • a newly changed beneficiary where the account details on an old payee were edited days before;
  • urgency expressed by the customer to the bank's own staff, which is the same pressure the fraud applies internally;
  • for a business customer, a first-ever international wire, which deserves a call rather than a straight-through payment.

On the receiving side the bank sees the other half, and this is where an institution can stop the money moving on. A receiving account that was opened recently, has little history, receives a large credit out of keeping with its stated purpose, and then disperses it quickly in smaller amounts or onward abroad, is the standard mule pattern. The Singapore case in the previous lesson turned on exactly this: part of the money had already gone onward immediately on arrival.

This is the point where fraud and anti-money laundering work meet, and where a divided organisation loses money. The sending bank sees an unusual payment; the receiving bank sees an unusual credit; between them they hold the whole picture and separately hold only half of it. The red flags published for this fraud remain useful precisely because they are drawn from reports filed by both sides.

A caution on expectations. These signals produce false positives in volume, because legitimate businesses also acquire suppliers and change banks. The aim is not to block the payment but to earn the right to make a telephone call before it leaves. A bank that cannot place that call quickly has a detection capability it cannot use.

Lesson 6 of 9

After it happens

The hour after discovery decides the outcome, so the sequence should be written down before it is needed.

The calls, in order. The paying bank first, to ask for a recall and to have it contact the receiving bank. Then the police or the national fraud reporting body, because some freezing routes only open through law enforcement. Then, if the payment crossed a border, say so at once: the FBI's figures show international freezes are both fewer and slower than domestic ones. Someone should hold this list, with the telephone numbers filled in, and it should not live only in the email system that may be compromised.

Then the reporting. A regulated institution that handled the payment will usually have a suspicious transaction or activity report to consider, and FinCEN's advisory exists partly to shape that reporting. The victim business has its own duties, which may include notifying a data protection authority and its insurer. These are deadlines, not formalities, and a prepared list makes them far easier to meet.

Assume the email account is still compromised. If the criminal was reading messages, they are probably reading the ones about the fraud. Move the conversation off that system, force a password reset, check for forwarding rules and for mailbox rules that hide replies, and review what else the account could reach.

Remember that not every BEC is after money. FinCEN's advisory highlights a data-theft variant, in which a compromised account is used to obtain employee or customer records rather than a payment. The request looks administrative, the loss is personal data, and the consequences run through data protection law instead of the payments system.

Then the part most organisations skip. BEC is a process failure, and the person who made the payment followed a process that permitted it. Treating the incident as an individual's mistake guarantees a repeat, because the process that allowed it survives. The review should ask which control was absent: was there no out-of-band verification, no second approval on a details change, no delay on a first payment, no number on file to call?

Finally, tell people. Organisations that circulate what happened, in plain terms and without naming the individual, see the next attempt reported in minutes rather than days. The fraud relies on each recipient believing the message is ordinary, and that belief is the only thing an organisation can reliably take away from it.

Lesson 7 of 9

How the criminal prepares

The payment request is the last step. Most of the work happens before it, and almost all of that work is observation rather than technique.

Learning the relationship. Before anything is sent, the criminal needs to know who pays whom, who approves it and what the two parties call each other. A compromised mailbox supplies all of it: past invoices, the finance team's names, the tone the supplier uses, the day of the month payments run. FinCEN's account of property fraud makes the same point about public information, where the parties and sometimes the sums are a matter of record.

Choosing the address. Two routes, with different consequences. Sending from the real account is the harder version to spot, because every technical check passes: the address is genuine, the history is genuine, and the message may sit inside a conversation that both parties have been having for weeks. The cheaper route is a lookalike, which FinCEN describes as an account closely resembling a known counterparty's, altered by adding, changing or deleting one or more characters. A doubled letter, a swapped letter pair, or a different ending to the domain will survive a glance at a small screen.

Hiding the reply. Where a real mailbox has been taken, the criminal needs the genuine owner not to see the answers. Rules that move or delete messages matching the supplier's name, or forward them elsewhere, do that quietly, which is why an incident review that does not look at mailbox rules has not finished.

Waiting for the moment. The request lands when it is least likely to be questioned: against a real invoice already in the system, during a known absence, at a quarter end, or while a genuine change of banking is under way. Urgency is then supplied as a reason not to check.

What this means for defence. Every one of those steps assumes the paying organisation will treat the message as its source of truth about where money goes. A process that keeps that truth somewhere else, in a supplier record that only changes through a verified route, defeats the whole sequence regardless of how good the imitation is.

It also explains why the better imitations cannot be spotted by reading carefully. Staff are regularly told to look for poor spelling and odd phrasing, advice that was never strong: the National Cyber Security Centre's own case study of a bulk invoice-themed campaign describes the messages as well written, with good spelling and grammar. In this fraud the words are often the supplier's own, copied from a real thread, and the only reliable tell is the one thing the criminal cannot copy, which is the account number you already hold on file.

Lesson 8 of 9

The institution's own exposure

A bank or payment firm meets this fraud in three different roles, and conflating them is a common planning error.

As the payer. An institution has suppliers, a payroll and premises like any other business, and FinCEN's finding that around half of BEC aimed at financial institutions impersonated the chief executive or president is a statement about institutions as victims, not as conduits. The earlier controls apply internally, and the institution has no special protection from them.

As the sending bank. Here the customer gives a valid instruction and the institution executes it. Its exposure is not usually legal, for the reason set out earlier: the payment was authorised because the payer consented. It is reputational and operational. The customer who loses a large sum on a payment the bank could have queried will say so publicly, and will be right that the bank saw the only unusual payment in that account for a year.

As the receiving bank. This is where an institution holds real power and real risk. The account that receives the proceeds is a customer of someone, it was opened through someone's onboarding process, and it moves the money on through someone's payment rails. Weak onboarding in one bank is the engine of fraud losses in every other bank, which is why mule account controls belong in the fraud programme and not only in the anti-money laundering one.

The reporting that follows. FinCEN's advisory exists in part to shape suspicious activity reporting on this fraud: it asks institutions to report, and the typologies and red flags it lists are drawn from what institutions reported before. That is a loop worth understanding. The quality of the next advisory, and of the red flags your own analysts will work from, depends on the detail institutions put in reports today.

And the obligation to act on what is reported. An institution that receives a freeze request about an account it holds is the fastest-moving part of the whole system. The FBI's domestic figures, 2,651 complaints and 469.1 million dollars frozen in 2024, are the sum of many individual receiving banks acting within hours. The international figures, 369 complaints and 92.5 million dollars, show what happens when that chain has to cross a border and a time zone.

The practical conclusion is unglamorous: have a named route for freeze requests, make it answerable out of hours, and test it.

Lesson 9 of 9

Running a programme against it

Everything above becomes a programme when somebody owns it, it is tested, and it is measured. Without those three things an organisation has a policy and a hope.

Ownership. BEC falls between departments: finance makes the payment, information security holds the mailbox, procurement owns the supplier record and fraud owns the loss. That division is what the fraud exploits. One named owner, with the authority to change the payment process rather than merely warn about it, is the first control.

A written process for the one decision that matters. Who may change a payment detail, by what route it is verified, who approves it, and what happens to the first payment afterwards. It should be short enough to be read, specific enough to be audited, and should name the number to call.

Testing, not training alone. A simulated request, sent with the finance team's agreement and using the organisation's own supplier names, tells you what the process does under pressure. Measure what matters: not whether people spotted it, but whether the payment would have left. A test that someone reports in five minutes and a test that nobody reports but the control stopped anyway are both passes, and they are different passes.

Measurement. Four numbers are worth a board's attention: how many payment-detail changes were requested, how many were verified by the defined route, how long a first payment to a new account is held, and how long it takes from discovery to the first call to the bank. None of them is a count of attempts, which measures the criminals' effort rather than your process.

Rehearse the hour after discovery. The call list should be exercised once, with the numbers actually dialled. Organisations that do this find out that the fraud line has changed, or that the emergency contacts live only in the email system that may be compromised.

Join it to the anti-money laundering side. Where fraud and anti-money laundering teams do not speak, the institution holds both halves of a case and sees neither.

Finally, be honest about residual risk. No set of controls reduces this to zero while businesses accept instructions from outside and pay large sums by transfer. The purpose is to make the loss small, recoverable and reportable rather than large, final and surprising. An organisation that can say how much it could lose on a single payment tomorrow, and how quickly it would know, is in good shape.

Knowledge check

Thirty questions

Answer all thirty questions, then check your answers. You need 27 out of 30 to pass and receive a certificate. If you score less, you will see which answers were right and wrong, and then go through the course again before you retake the check. Your answers, progress and times are kept only in this browser.

Sources

The official documents this course relies on. Laws and guidance change, so check the current version.

  1. 2024 Internet Crime Report (business email compromise, and the Financial Fraud Kill Chain) · Internet Crime Complaint Center, US Federal Bureau of Investigation
  2. Advisory FIN-2019-A005: Updated Advisory on Email Compromise Fraud Schemes Targeting Vulnerable Business Processes (16 July 2019) · Financial Crimes Enforcement Network, US Department of the Treasury
  3. Advisory FIN-2016-A003: Advisory to Financial Institutions on E-mail Compromise Fraud Schemes (6 September 2016) · Financial Crimes Enforcement Network, US Department of the Treasury
  4. Directive (EU) 2015/2366 on payment services, Article 64 (consent and withdrawal of consent) · EUR-Lex
  5. Phishing attacks: defending your organisation · National Cyber Security Centre, United Kingdom