In many ways, voice biometrics promised to be a sort of magic bullet for fraud risk among contact centres.
Just as physical fingerprints can be used to verify identity before allowing access into a physical space, voice biometrics can act as a protective safeguard during telephonic conversations. What’s more, nearly no human intervention is required for auto-identity verification through voice biometrics. However, there are a number of considerations to keep in mind before implementing voice biometrics technology as a default, 100% reliable security mechanism at your content centre – let us explore these pros and cons in more detail.
3 Undeniable Benefits of Voice Biometrics
The voice biometrics market is growing at an impressive pace, expected to reach $4.9 billion by 2027, compared to less than $1 billion in 2019. This growth is propelled by the following benefits:
- A person’s voiceprint is extremely hard to spoof – In an identity theft case, a fraudster can get hold of a customer’s date of birth, address, and unique information like their mother’s maiden name or the name of their first pet. However, every individual has a distinct voiceprint, which is far harder to obtain or mimic
- Its convenience helps the quality of CX – Voice-biometrics-based authentication is more convenient for the customer than physically entering a password, remembering answers to secret questions, etc. There is no dependence on memory or recall, as your voice itself acts as the identifier
- It is an excellent candidate for automation – Voice biometrics can be used to automatically obtain a person’s approval after verifying their identity. For instance, a live agent doesn’t need to ask a customer for consent before recording a call – an automated IVR can request, while voice biometrics verifies that it was indeed the customer who provided their consent
As a result, there is a lot of interest and investment around voice biometrics, with nearly every technology company opening a patent in this field.
3 Potential Disadvantages of Voice Biometrics
Right at the outset, one should remember that voice biometrics isn’t a mature technology. It relies on sophisticated AI algorithms, trained on comprehensive datasets, and tested in real-world scenarios – which can be difficult to achieve. This can lead to the following pitfalls in biometrics implementation:
- There is a risk of discrimination and racial bias – As mentioned, voice biometrics algorithms
(especially those used for identification and not verification), have to be trained on comprehensive datasets, comprising a diverse range of human voices. But studies suggest that the dataset used for training most mainstream voice technologies might be racially skewed, which makes the AI better at recognising some demographics than others - Companies must be extremely sensitive about privacy and consent – Voice is innately personal and not every customer will be comfortable with sharing their voice data. Enforcing data privacy laws like GDPR can be problematic, as customer voice samples are relatively easy to collect. Already, the Chinese government has come under scrutiny for potentially breaching privacy rights and collecting voice pattern samples to build a national database
- Voice deep fakes are possible – Finally, audio deep fakes are becoming increasingly common and may be able to fool the Ai into believing the audio sample’s veracity