- BEC is defined by impersonation and a fraudulent payment request, not by the absence of malware: many messages are plain text, but campaigns also use credential phishing, malicious links and account takeover
- The four main attack types are CEO fraud, invoice fraud, account compromise, and attorney impersonation
- DMARC at p=reject blocks exact-domain spoofing at receivers that enforce it, but not lookalike domains, display-name impersonation or mail from a compromised genuine account
- MFA greatly reduces password-based account takeover; phishing-resistant methods such as passkeys close more of the remaining gap
- Independent confirmation of bank-detail changes, using contact details from a previously validated record, is the strongest process control
- If a fraudulent payment goes out, call your bank immediately; recovery is possible but never guaranteed
What BEC is and how it relates to phishing
Business email compromise is financial fraud carried out through email and related channels. An attacker impersonates someone the target trusts, such as your CEO, a long-standing supplier, or your solicitor, and persuades someone in your organisation to send a payment or change the bank details on file. The FBI defines BEC by that combination: impersonation through compromised or spoofed communication, and an unauthorised transfer of funds, achieved through social engineering or computer intrusion. The request usually arrives at a plausible moment, often a Friday afternoon or mid-payment cycle.
BEC and phishing are not mutually exclusive. Many BEC messages are pure social engineering: plain text, no attachment and no link, which gives filters that look for malicious content little to work with. But many BEC campaigns start with phishing. A credential phishing page or a malicious link gives the attacker access to a real mailbox, and that account takeover is then used to send payment requests from inside a genuine conversation. The practical difference is the goal. Mass phishing usually aims to harvest credentials or install malware at scale; BEC aims at one payment, prepared with research and timed to fit a real business process. Email security tools catch part of BEC, particularly spoofed domains and known phishing infrastructure, but the final payment request often looks entirely legitimate. That is why financial process controls carry so much weight.
The FBI's Internet Crime Complaint Center (IC3) recorded US$3.05 billion in BEC losses across 24,768 complaints in its 2025 Internet Crime Report, published in 2026. These are complaints reported to IC3, mostly from the United States, not a global estimate. Many organisations never report an incident, so actual losses are higher.
The four main attack types
BEC incidents follow recognisable patterns. Understanding the format of each one shapes both your technical controls and the process changes your finance team needs to make.
CEO fraud
The attacker spoofs, imitates or compromises the CEO's email address and sends a message directly to a finance team member or accounts payable. The message requests an urgent transfer, often framed around a time-sensitive acquisition, a confidential deal, or a legal requirement. The urgency discourages the recipient from following normal approval procedures. The CEO is "travelling" or "unavailable to take calls." The request often instructs the recipient not to discuss the transfer with colleagues.
CEO fraud works because authority and urgency together suppress verification behaviour. An employee receiving a direct request from the company's most senior person, marked confidential and time-sensitive, faces social pressure that makes pausing feel obstructive. Attackers research the CEO's travel schedule, writing style, and typical language from LinkedIn posts, press releases, and out-of-office replies. Some now add a phone call or voice message using a cloned voice; our article on voice deepfake CEO fraud covers that variant, and the deepfake and identity fraud guide covers video and identity manipulation more broadly.
Invoice fraud
The attacker impersonates a supplier your organisation pays regularly. The message notifies your accounts team of a change in bank details and asks that future payments go to the new account. The impersonation targets suppliers you pay large amounts to, identified from publicly available company information, procurement announcements, or construction contract notices.
Invoice fraud is among the most damaging BEC variants because payments go out under routine circumstances, processed by staff doing their normal jobs. There is no urgency trigger and no unusual request. The only anomaly is the changed bank details, and that anomaly arrives with a plausible explanation: an accounting system migration, a new banking partner, or a company restructure.
Account compromise
In account compromise attacks, the attacker gains genuine access to a legitimate email account, at your organisation or at a supplier. Common routes are credential phishing, adversary-in-the-middle phishing kits that capture the session as well as the password, malicious app consent, and passwords reused from a data breach. From inside a real account, the attacker may monitor email for days or weeks, identifying payment relationships, approval workflows, and the language of the account owner, and often creates inbox rules that hide or forward replies. When a real invoice or payment request arrives, the attacker intercepts it and changes the bank details.
This variant is hard to spot because the email comes from the real account, passes SPF, DKIM and DMARC, and sits inside an authentic thread. Sender checks and external banners do not help. The main defences are independent verification of payment changes and preventive controls that make accounts harder to take over and faster to contain.
Attorney impersonation
The attacker poses as a solicitor, notary, or legal adviser handling a confidential transaction on your behalf. This variant targets business owners and executives directly, often around genuine high-value events such as a property purchase, a merger, or a legal settlement. The "lawyer" provides payment instructions for a deposit or settlement amount and stresses confidentiality to deter the target from discussing the transfer with others.
Attorney impersonation exploits the deference many people extend to legal professionals and the genuine confidentiality obligations that surround legal transactions. Targets often receive these messages when they are already in the middle of a stressful transaction and expecting payment instructions to arrive.
Most BEC attempts share a handful of structural features: a request for financial action, time pressure, a reason to bypass normal process, and an instruction to keep the communication private. Any email that combines two or more of these elements deserves verification before action, regardless of how legitimate the sender appears.
Why financial services and accountancies are prime targets
BEC attacks follow the money. Financial services firms, accountancy practices, and wealth managers attract disproportionate attention for three reasons.
First, they process large transfers as a matter of routine. A payment of £250,000 that would trigger multiple approval steps at a manufacturing company is a normal daily transaction for an accountancy firm managing client funds. The threshold above which finance staff pause and verify is higher, and attackers know it.
Second, they hold money on behalf of third parties. A law firm holding client account funds or an asset manager executing investment instructions moves money at client request without the same organisational controls that apply to internal payments. An attacker who convinces a conveyancing solicitor that a client has changed their completion account can intercept an entire property purchase.
Third, they are rich in public information. Regulatory filings, Companies House records, LinkedIn profiles, and client announcements make it straightforward to identify who handles payments, who the senior partners are, and which client relationships are active. This shortens the reconnaissance needed before an attack.
Regulated firms also face a compliance dimension. Depending on the firm and the incident, a BEC loss can raise notification questions with the FCA, with the ICO if personal data is involved, and with clients and insurers. Each obligation has its own test, covered in the response section below. The reputational and regulatory exposure compounds the direct financial loss.
BEC does not break into your system. It walks in through the front door by convincing someone to hold it open.
How attackers build credibility
A BEC email that reads like a generic fraud attempt fails. Attackers invest real time in building the context that makes requests feel routine.
Domain spoofing and lookalike domains
The simplest technique is display name impersonation: the attacker sets the display name to "Sarah Mitchell, CFO" while the actual sending address is a throwaway webmail account. Many email clients, especially on mobile, show the display name prominently; the actual address takes a deliberate tap or click to inspect. Without DMARC enforcement on your domain, exact-domain spoofing is also possible. The attacker sends as sarah.mitchell@yourcompany.co.uk and the email arrives appearing to come from your own domain.
Lookalike domains take more effort and are not affected by your DMARC policy at all, because they are a different domain. The attacker registers yourcompany-accounts.co.uk, yourcompany-invoices.co.uk, or yourcornpany.co.uk (with a substituted character). These domains can publish their own SPF, DKIM and DMARC records and pass authentication, so the receiving mail server has no technical reason to reject them.
Email header manipulation
Reply-to manipulation lets attackers use a convincing from address while routing replies to an address they control. The email appears to arrive from ceo@yourcompany.co.uk, but if the recipient clicks Reply, the message goes to ceo-urgent@gmail.com. Recipients rarely inspect the reply-to header before responding. Over several exchanges, the conversation moves entirely to the attacker-controlled address.
OSINT and LinkedIn research
Open-source research takes a motivated attacker little time. LinkedIn shows the names and roles of finance staff, who reports to whom, and recent promotions. Job postings reveal the software your accounts payable team uses. Press releases name your key suppliers. A CEO's recent conference talk gives the attacker their vocabulary and communication style. Out-of-office replies confirm travel dates and provide the name of the colleague who handles things in their absence, who then becomes the next target.
Five techniques compared
Technical controls that reduce exposure
Technical controls do not stop BEC on their own, but they shut down some techniques and make others harder and more visible.
DMARC, DKIM, and SPF
SPF (Sender Policy Framework) lists the mail servers authorised to send email for your domain. DKIM (DomainKeys Identified Mail) adds a cryptographic signature that receiving servers can verify. DMARC (Domain-based Message Authentication, Reporting and Conformance) builds on both: a message passes DMARC when SPF or DKIM passes and the domain it authenticates aligns with the domain in the visible From address. The DMARC policy tells receivers what you want done with mail that fails. p=none is monitoring only and requests reports. p=quarantine is an enforcement policy that asks receivers to treat failing mail as suspicious, typically by sending it to spam or quarantine. p=reject asks receivers to reject it.
For domains that send email, publish SPF and DKIM for every legitimate sending service, start at p=none to collect reports, and move to p=quarantine and then p=reject once you have confirmed that all legitimate senders (your mail platform, CRM, invoicing, marketing and payroll services) pass with alignment. Enforcing before that can block your own invoices. For domains that send no email, publish a null SPF record (v=spf1 -all) and a DMARC record with p=reject. DKIM is not needed for a domain that sends nothing; an empty DKIM record is optional. An unprotected parked domain is an easy spoofing resource.
Check your current DMARC status with a tool such as dmarcian's DMARC inspector. Many organisations that believe DMARC is in place discover it is still at p=none, which monitors but asks receivers to block nothing. Even at p=reject, the result depends on the receiving server applying your policy, and DMARC does nothing against lookalike domains, display-name impersonation or mail sent from a compromised genuine account.
MFA and account protection
Multi-factor authentication on Microsoft 365, Google Workspace, or any email platform greatly reduces password-based account takeover: a password obtained through phishing or a credential dump is no longer enough on its own. It does not close the account compromise vector. Adversary-in-the-middle phishing kits capture the session token after the user completes MFA, stolen session cookies can be replayed, malicious OAuth apps can gain mailbox access through user consent, and a compromised device exposes whatever its user can reach. Our article on how attackers bypass MFA covers these techniques.
Prioritise phishing-resistant methods such as passkeys and FIDO2 security keys, starting with finance staff, executives, and administrators. Combine them with Conditional Access policies based on device compliance and sign-in risk, sign-in and session monitoring with fast token revocation, restrictions on user consent to third-party apps, and an account recovery process that cannot be bypassed with a convincing call to the help desk.
Shared mailboxes need a different approach. In Microsoft 365, block interactive sign-in for shared mailbox accounts and give access through named, delegated users who sign in with MFA. Inventory service accounts, limit them to the permissions they need, and restrict them with Conditional Access where possible. Our Microsoft 365 security hardening guide covers these settings.
Email banner warnings for external senders
Configure your email platform to tag messages from outside your organisation, with a banner or with the external sender tag built into Outlook. It gives recipients a useful cue to check the sender address, and it is available in Microsoft 365 and Google Workspace at no additional cost. Treat it as an auxiliary control. It does not eliminate display-name impersonation, it does not flag messages from a compromised internal mailbox or a genuine supplier account that has been taken over, and users stop noticing a banner that appears on most of their mail.
Monitoring for lookalike domain registrations
Services such as DomainTools, Brandwatch, or DMARC reporting platforms surface newly registered domains that closely resemble yours. A daily alert for domains containing your company name gives you visibility before a campaign launches. Some organisations also pre-emptively register common typosquat variants of their own domain, taking those names out of attackers' hands.
Process controls
Technical controls reduce attack surface. Process controls catch the requests that get through anyway, and for BEC they carry more weight than for almost any other threat.
Verification callbacks for payment changes
Every request to change bank details for a supplier, client, or employee should trigger independent confirmation before the change takes effect, whatever the amount. Call a number taken from a previously validated record, such as the supplier onboarding file or the contract, not from the email requesting the change or any contact details it supplies.
This is the strongest single process control, but not a guarantee. It fails if the supplier record was already altered by an earlier fraudulent change, if the contact you call has been compromised too, or if staff skip the procedure under pressure. Done consistently, with contact details from a validated record, it significantly reduces the risk of invoice fraud and of payment redirection from compromised accounts. It needs a written policy, training, and a culture in which making the call is expected rather than seen as obstructive.
Dual authorisation and segregation of duties
Require two separate people to approve payments above a threshold, with both approvals logged. Size the threshold to your organisation: for many SMEs, £5,000 is a reasonable example to start from, not a rule. Apply the same principle to supplier master data. A change to a supplier's bank details should be entered by one person and approved by another, separately from the person who releases payments. For higher-risk payments, such as large transfers, first payments to new beneficiaries, and urgent requests outside the normal cycle, combine dual authorisation with confirmation through a channel other than the original email.
CEO fraud works by framing a request as confidential so that it bypasses the second approver. A written policy stating that no instruction from an executive overrides dual authorisation removes that pressure point.
Payment instruction confirmation windows
Consider a hold period, for example 24 hours, for first payments to new beneficiaries, payments to recently changed bank details, and payments above a defined threshold. The period is an example; size it to your cash-flow needs. A hold gives time for independent confirmation before funds move. Many business banking platforms can flag new payees or high-value transactions for manual review. Configure these rules and test them periodically.
Finance teams and executive assistants need scenario-based training, not annual compliance slides. Walk them through real BEC email samples, show them how to inspect sender headers in Outlook and Gmail, run a simulated BEC exercise targeting a plausible payment request, and measure how many people would have processed the transfer. The result sets your baseline and identifies individuals who need additional coaching.
What to do if you are hit
Act immediately. Recovery chances fall as funds move on from the first receiving account, and no fixed window guarantees a return. The moment anyone suspects a payment went to a fraudulent account, start the steps below in parallel.
Call your bank immediately. Do not rely on email or a portal request. Call the bank's fraud line using the number from your account documentation, ask it to attempt a recall, and ask it to contact the receiving bank. A recall request does not necessarily reverse a payment that has already settled; the outcome depends on the receiving bank and on funds remaining in the account. The Faster Payments reimbursement rules for authorised push payment fraud cover consumers, micro-enterprises and small charities, so most businesses should not count on reimbursement.
Report the crime. Report to Report Fraud, the national service that replaced Action Fraud, at reportfraud.police.uk or on 0300 123 2040, and keep the reference number for your bank and insurer. Notify your cyber insurer early and check the policy's notification window; some policies only cover social engineering losses under a specific extension.
Assess reporting obligations separately. A financial loss on its own does not trigger every obligation. If personal data was exposed, for example through a compromised mailbox containing client details, assess the incident under UK GDPR: report it to the ICO within 72 hours of becoming aware if it is likely to result in a risk to people's rights and freedoms. FCA-regulated firms should assess whether the incident needs notifying to the FCA, and contracts with clients and insurers have their own notification terms.
Response by workstream
After the immediate response, establish how the attack succeeded. Was it a compromised account, a lookalike domain, or a process failure? The answer determines what you fix. A post-incident review within days, while detail is fresh, should produce a written findings report and a remediation plan with named owners and deadlines.