A fraudulent message using a relative's voice or a manager's video works because it creates a context where there seems to be no time to check, even if the fake is imperfect: an accident, a hospital, a collapsed deal, or a secret security operation. The stronger the emotion and urgency, the less attention remains for odd facial movements and audio defects.

The Bank of Russia recommends taking time before making a transfer, calling the person yourself, and asking a personal question. This matters more than any list of visual signs. Generation quality changes, while an independent verification channel remains effective.

Verify the action rather than identifying the fake

A voice, photograph, account name, and even a video call are message content, not proof of identity. Verification must happen through a different channel that you initiate. End the incoming call and dial a saved number. Write in the old family chat. Contact a colleague through the corporate directory, rather than using the contact in a new email.

A five-step protocol

  • Pause: do not transfer money or disclose a code for at least ten minutes.
  • Independent channel: contact the person using a number or address you already had.
  • Personal question: ask about a fact absent from public profiles and conversations with the new account.
  • Second person: involve a relative, colleague, or manager for a large transfer.
  • Check payment details: compare the recipient with previous payments and agreements.

A secret family word is useful if it is kept out of conversations and changed periodically. But do not make it the only defense: any secret can leak. Reliability comes from combining a pause, an independent call, and a second person.

For a small company: the two-confirmation rule

In business, a deepfake often disguises itself as an exception: a manager supposedly asks for an urgent change of payment details or an invoice payment outside the normal process. The defense is a rule against changing the payment route through a single message, rather than voice-recognition training.

Set an amount above which two approvals are required. Confirm changes to bank details through a known contact at the counterparty. Do not use one messenger as both the request channel and the confirmation channel. Management must publicly support the rule: staff are not punished for delaying a suspicious payment.

Protect the material used to create fakes

It is impossible to hide your voice and face completely, but you can reduce the attack surface. Limit long public speech recordings, check the privacy settings of old profiles, delete unused accounts, and enable multifactor protection. Email is especially important: it is used to recover other accounts and gather context for a convincing story.

  • A unique email password and a password manager.
  • Multifactor authentication, preferably through an app or hardware key.
  • New-login notifications and backup codes stored outside the phone.
  • A monthly review of active messenger sessions.
  • Minimal public details about payments, travel, and company structure.

If the transfer has already been made

Contact the bank immediately using its official number and report fraud. Save messages, numbers, payment details, times, and screenshots — do not keep arguing with the attacker. Change the password from a clean device, end active sessions, and warn contacts if the account may have been compromised.

The key everyday habit is simple: urgency creates a need for verification instead of canceling it. When the protocol is known in advance, “don't tell anyone and do it now” becomes a signal to stop rather than a command.

Rehearse verification before a real call

A protocol only works when people remember it under pressure. Once, act out a safe scenario with your family or team: a message comes from an unknown number, the voice sounds familiar, and the caller asks you not to call back. The person should say the next step aloud, find the saved contact, and involve a second participant. The rehearsal takes five minutes and exposes practical gaps: the number is not saved, the relative does not answer, or nobody knows a backup contact.

Agree in advance that a verification call is not a sign of distrust or rudeness. A fraudster wins when a victim is afraid of offending a loved one or manager. A normal response is: “I will help, but first I will complete our verification.” Do not argue about video quality or ask them to turn their head in a particular way: technical signs change, while an independent channel remains a clear rule.

  • For a family: two known numbers, a backup relative, and a phrase that signals a stop.
  • For a team: a payment threshold, two approvers, and the bank's official contact.
  • For a counterparty: payment details change only after a callback using the previous contact details.
  • For yourself: never install an app or enable screen sharing at the request of an incoming caller.

Review the incident without blame

Even if the attempt failed, record how the attacker knew names, job titles, travel plans, or the amount. Check public profiles, email leaks, forwarded documents, and compromised accounts. If someone has already sent money, the priorities are the bank, blocking access, and preserving evidence, rather than looking for someone to blame in the chat. Shame makes people hide the first minutes of an incident, even though those minutes matter most.

Then change the specific element that failed: add a backup contact, lower a limit, enable MFA, delete an unnecessary public recording, or make payment-details confirmation mandatory. Do not try to ban all voice messages and video calls. The goal is to prevent one convincing fragment from triggering a transfer or granting access on its own, rather than detecting every fake.