Passwords are passe; biometrics is the future of banking

Banks are now adopting voice recognition software to authenticate customers

You call up your bank. The customer executive asks you for your name and other details. As you speak, a software in the background analyses your voice and authenticates that the caller is the real account holder. It can then let the user carry out tele-banking transactions.

This is something you may encounter soon as more and more banks world over are using biometrics to confirm the identity of customers when they dial the call centre. For one, it saves the hassle of entering the account details and passwords and it also shortens the call duration.

In India, ICICI Bank has taken the lead to introduce voice recognition software. World over, Abu Dhabi Commercial Bank (ADCB) of UAE, Mexico’s Banco Santander Mexico, and Barclay’s are among the banks that have been using it.

Since April this year, ICICI Bank started creating a bank of customers’ voice samples. Whenever a customer calls from a registered mobile phone and talks for 60 seconds, the software captures the voice pattern and stores it. Next time onwards, whenever the person dials the call centre, he gets an option to choose voice authentication. Once he opts for this, the phone is transferred to an executive. While the person talks for 10 seconds giving out details, the voice patterns are analysed and matched. This is second level of authentication. The first is the registered mobile phone. If the customer does not call from the registered number, he will need to resort to the traditional methods of verification, that is, account number and PIN.

Using this, a person can carry out telephone banking transactions such as cash transfers to registered users, bill payments to registered vendors, recurring deposits and so on.

ICICI Bank has introduced the software in partnership with a US-based firm that uses 100 parameters to analyse the voice.

Experts, however, have a word of caution. “The combination of speaker verification with an alphanumeric password is certainly more secure,” says Richard Stern, professor, Department of Electrical and Computer Engineering, Carnegie Mellon University. His primary area of teaching and research is automatic speech recognition and related technologies. 

The system does not have a 100% success rate. The most respected speech recognition evaluation, run by the US National Institute of Science and Technology, shows that the best system falsely accepted 0.2% of the speakers. And it incorrectly rejected 5%. This brings the best accuracy of such systems to 98%.

To use this software, a person also needs to be in an environment where the background noise is not too high. If a customer uses it while the TV is loud in the background, or in a stadium, or while standing close to people talking in loud voices, the chances of software failing to authenticate the person is high.

Going forward, experts see more financial institutions adopting it. “I do see voice recognition in some form being used increasingly as the technologies become more accurate and more robust,” says Professor Stern. “They (speaker recognition systems) provide an additional layer of security for the financial institution with no further significant encumbrance on the part of the user.”

Compared to other biometric technologies such as fingerprint and iris scan, speech recognition is extremely easy to deploy, as only a microphone is needed as an input device, and these are commonly part of every phone and every computer now. It is also cheaper than other methods.