It would seem we’re all finally getting the message: we must be more careful with our identities. As examples of security threats increase, such as Governments misplacing citizen data, and even TV presenters being caught out when over-confidently publishing bank details in the newspaper, we’re realising how important it is to take care of our personal information.
This is affecting a number of daily activities. We’re on our guard when taking money out of cash points. We’re wary of waiters taking our cards away to authorise a payment in a restaurant. We place a watchful eye over the increase in online fraud, and when it comes to handing out our bank account details to an unknown person over the phone – well, how do we know they can be trusted?

Yet as this mist of shrewd suspicion descends over us, it puts doubt into people’s minds over other matters; matters which really needn’t worry them at all. This includes a variety of fantastic new technologies for businesses, which are actually designed to help them and their customers avoid fraudulent activity – such as speech recognition and speaker verification.

HOLLYWOOD HORROR STORIES

You’re probably already familiar with the use of speech recognition in business. The idea is that a person calls up a company’s contact centre and – rather than speaking with a real agent – has a conversation with an automated voice, perhaps to buy something or to make a general service enquiry. This allows agents to focus on customers who really do need personal assistance.

Particularly for customer transactions with large organisations such as banks, modern speech technology has the added benefit of being much more secure than a person-to-person call. Where the customer would typically verify their identity by handing over their postcode, date of birth or mother’s maiden name to the agent – most of which, let’s face it, could potentially be found nowadays on internet social networking sites such as Facebook and MySpace. Speaker verification compares the voice pattern of the caller’s utterance with the verified voice print the caller made at the time of enrolment. It does this using a biometrics engine but then uses business rules to determine whether it responds with a pass or fail. What’s more (and without wishing to pour any scorn on the agent workforce), this means there’s no risk of an agent misappropriating those details for their own ends.

On the face of it, this technology should be hugely appealing to businesses. But gaining the trust of both businesses and users is difficult, not least because Hollywood has vilified it. Thanks to Mission: Impossible and other films like it, we’ve been led to believe that one’s voice can easily be recorded, allowing evil-doers (and anti-heroes) to trick speech recognition systems in order to hack into safes, computers, bank accounts and offices at the mere click of the ‘play’ button.

But this is really very unlikely indeed.

I’M SORRY DAVE, I’M AFRAID I CAN’T DO THAT

This is because modern speech systems will use varying levels of authentication throughout a voice transaction, depending on the customer’s request. If someone asked, for instance, to transfer thousands of pounds into an offshore bank account, this would trigger a higher level of authentication that makes it much harder for the caller’s voice to be accepted.

Instantly, the application becomes more sensitive to environmental elements such as background noise, becoming quicker to reject the caller if the voice is unclear. What’s more, it also asks the caller to repeat critical utterances such as account names and numbers, to ensure the caller is real. Thus a recording that delivers an identical voice pattern would result in an instant rejection. And your money would remain safe.

Without doubt, most modern speech systems are much cleverer than many of us give them credit for. They can recognise and accept you if you have a cold and your voice is a little husky. They can even distinguish the voice patterns of twins. In fact, many systems perform regular updates to your voiceprint in the background, during the call, to ensure voiceprint records are current – and without the caller being aware of it whatsoever.

It’s vital to remember, however, that security systems never rely on speech by itself to guarantee privacy. In all cases speaker verification has to be combined with other factors to create a solution that is genuinely, entirely, secure. Typically this requires callers to remember certain numbers from a ‘special date’, for instance, that only they will know. This ‘multi-factor’ authentication is absolutely critical.

CONCLUSION

The great thing about speech-based authentication, though, is that these other security elements become much less intrusive. For callers, the system is easier to use as these more traditional authentication technologies recede into the background; still vital, yet ancillary.

And that’s part of the beauty of this technology. It’s designed to work around the caller. It doesn’t require them to remember their bank balance in order to be authorised, or to recall the date of their last purchase. All the user needs is reassurance that they can trust the technology in the first place. And after that, they only need the one thing they’re guaranteed to have with them: their voice.

To find out more about speech recognition solutions from Nortel, visit www.nortel.com/take-command