The anatomy of a scam call
The security team, an urgent debit, the code from an SMS — that script does not fool the system.
What the fraudster says and what the system sees
The pattern of the call is always the same: claimed to be the bank, frightened them with a debit, asked for the code, pressed with a deadline. The analytics recognises it by the words rather than the number — and flags the conversation before the customer says the code.
The call is staged and was recorded by actors. The names and numbers are invented.
- A request for the code from an SMSA bank never asks for the code. The analytics recognised a request for a one-time password — a level two alert.
The customer did not get as far as saying the code. The caller’s voice has been added to the fraud database.
We recognise the customer by voice — and the fraudster too
When a customer calls the bank, biometrics matches the voice against the stored print in seconds. When someone calls a customer in the bank’s name, the same engine matches the caller’s voice against the staff database and the fraud database.
- Match against bank staff0 of 412
- Match against the fraud database3 calls this month
- Number in the bank’s directoryNo
The voice database grows with every call
A fraudster’s voice does not change from one number to the next. A call flagged once adds the print to the database, and the next call from a new number is recognised in the first seconds.
1
The call is flagged by the analytics or the security team
2
The voice print is added to the database
3
The next call from any number is recognised within 5 seconds
- Customer
- D. Rakhimov, card ···· 4471
- Risk
- High: code requested, pressure applied
- The caller’s voice
- In the database, 3 matches
The security team finds out during the call, not from a complaint
The alert reaches the duty officer through the channel you choose: the call record, the transcript, the risk level and the action buttons.
What changes in the work
Where the numbers move once every conversation is reviewed instead of a sample.
Less time spent verifying identity: the customer simply talks while the voice is already being matched against the print.
Fewer losses to the schemes where the caller passes themselves off as a bank or insurance customer.
More attempts at deception are visible before the conversation reaches a transfer of money.
A list appears of known fraudsters: one voice links all of their attempts across every channel.
Still have questions?
How does the system listen to a fraudster’s call to a customer if it does not go through the bank?
It does not. The scenario on this page is a call through the bank’s telephony or a protected call in the bank’s app. For outside calls the system protects the other side: when a fraudster calls the bank posing as a customer, their voice gives them away.
What if the fraudster’s voice is altered by software?
Biometrics flags the signs of synthesised or altered speech and raises the risk level. A sign of synthesis is in itself a reason to check.
Is it lawful to store fraudsters’ voices?
Storing biometric data is governed by personal data law: we agree the procedure with your lawyer.
We will check your calls for signs of fraud
Send us recordings already established as fraudulent. We will show at which second the system would have raised the alarm.
