Voice authentication and
fraud detection

Prevent fraud and boost CX – use a customer’s unique voice pattern to instantly identify them during a call.
Request more info

Since the theft of personal information is more commonplace than it used to be, businesses cannot safely rely on using individual security questions to distinguish legitimate customers from fraudsters during voice interactions. At the same time, customers also expect faster interactions – security questions tend to be annoying, are often forgotten and may take a significant amount of time to go through and answer (increasing interaction cost). This creates a need for businesses to improve their authentication processes to not only enhance the customer experience but also to decrease their exposure to security risks.


  • Increase the likelihood of a faster resolution by eliminating the need for customers to answer security questions before they are able to get assistance.
  • Instant authentication reduces the customer effort associated with interacting with your business.
  • Voice authentication happens in the background of a call and in a few seconds, while security questions can take much longer to answer. This reduces the average handling time (and associated costs) significantly.
  • Reduce your business’ exposure to risk by proactively identifying potentially fraudulent activity before it occurs.

Key features

Passive enrolment

Customers do not need to repeat pre-defined phrases in order to enrol with the authentication application.

Mass enrolment

Use historic recordings to enrol customers and reap the benefits of biometric authentication and fraud detection immediately.

Known fraudster database

Most fraudulent activity is performed by repeat offenders. Create a fraudster database to flag known offenders before they even attempt to dupe agents.

Historical analysis

Identify past fraudulent activity and take corrective action. This is achieved by analysing previous recordings and performing retro-active authentication and fraud detection.

Configurable alerts

Using confidence scores, alerts can be delivered via email, or desktop pop ups to guide agents and their supervisors in real-time. You can then decide what protocol to follow should a customer show high, medium or low authentication confidence intervals.

Deliver confidence scores to 3rd party applications

This enables certain actions to be performed on a business application, such as rejecting a requested change to profile details if confidence scores are too low.

Include all calls or only a subset

Use rules to target specific types of calls for authentication. Authentication can also take place in parallel to other biometric comparisons.

Real-time or post process authentication

Depending on your requirements, authentication can occur in real-time or after the interaction has taken place.

Authenticate customers and agents

Ensure that you are not only speaking to the right customer but also that the agents are logged into the correct accounts and are who they claim to be.

Archive enrolment calls

Create a specific database to store the recordings used for enrolment. This will reduce risk if technology is changed at a later date.

How does voice authentication work?

Voice authentication relies on voice biometric data to recognise individuals based on measurable biological and behavioural characteristics. These characteristics are then used to create a unique digital pattern of your voice (called a voice print) which is used during calls to identify a specific individual.

Use of passive voice biometrics enables authentication and enrolment to happen silently and accurately in the background of a call. This occurs without interrupting the conversation or requiring any action from the customer.

The Verint voice authentication and fraud detection module alerts agents, supervisors or third-party applications of:

  • Identity confidence levels. This can be presented to the agent with a green, yellow or red indicator within a few seconds of a customer conversation. Organisational policies can then dictate the course of action based on each of the respective states. A security question might be asked with a yellow confidence level score, for example, whereas a green indicator could indicate that the agent can continue with the interaction.
  • Known fraudsters. If the customer is identified as a known fraudster with a confidence level above a set threshold, then the supervisor and/or agent can be notified, and the correct procedure followed.


According to a recent study, contact centre fraud has gone up by 350% in the past few years

Legitimate customers fail knowledge-based authentication questions 15% of the time

Security questions lengthen calls by 30-45 seconds, resulting in an increase in customer frustration, overall handling time and operational costs

Calculating ROI

Given the average authentication time of 30 to 45 seconds, conservative savings can be calculated by multiplying: (authentication time) x (amount of calls) x (agent burden cost).

Calculating ROI

Given the average authentication time of 30 to 45 seconds, conservative savings can be calculated by multiplying: (authentication time) x (amount of calls) x (agent burden cost).

Contact us now to get started