As technology evolves, so too must our approach to security. The era where “my voice is my password” is no longer the ultimate safeguard. With advancements in AI and biometrics, we’re stepping into a realm where multi-factor authentication and sophisticated algorithms redefine what it means to protect our digital identities.
If voice is no longer my password, then what is?
A straightforward solution is running text-independent, also called passive voice biometrics, in tandem with existing “my voice is my password” verification. This ensures that the same speaker who was verified in the first place is the speaker interacting with the call centre agent.
Whilst it is not foolproof, it is much more challenging and expensive for a fraudster to use voice cloning throughout interactions with human customer service agents than to spoof the initial “my voice is my password” prompt. This is especially true if you have customer service agents trained to be wary of voice-cloned calls. This is probably an adequate solution for low-risk, low-cost, regular customer service interactions.
In addition, asking the caller to repeat or respond to a random challenge can also help detect replay or voice clone attacks. If the repeated response is acoustically too similar to the initial response, this is highly suspicious. The one thing humans cannot do is say the same thing twice the same way.
These approaches are equally effective in automated speech recognition services. Here, the voice recognition command continuously checks each voice against the voice biometric, providing constant reassurance of the speaker’s identity throughout the session.
This is before you get onto implementing specialised algorithms tuned to detect acoustic anomalies in cloned speech.
All these approaches apply in the telephone call centre. However, as customer service evolves into a verbal and visual interaction, fused voice and face recognition comes into its own. Cloning a voice is one thing; cloning an image is another. But making a convincing clone of a speaking face is a much more demanding and far more expensive proposition.