Target detection

Organizations with high financial vulnerability can maintain watch lists of people suspected of fraud, known as targets. The voice of a person on an interaction is fed to the Voice Biometrics Engine. The engine compares the voice of the person with all the voiceprint models of targets from one or more watch lists. The score resulting from the comparison indicates the likelihood of target detection having occurred.

Using Recorder Analytics Rules for target detection

Target detection requires the following to be defined in each Recorder Analytics Rule:

  • Choosing which interactions you want to analyze.

  • Selecting the targets you want to detect by selecting one or more watch lists of targets. The Voice Biometrics Engine compares the person speaking against all voiceprint models from the selected watch list.

  • Defining the Minimum Score for the rule, which establishes the threshold for target detection. You can adjust the Minimum Score to control how frequently the Detected result occurs

Recorder Analytics Rule actions for target detection

During target detection, the results are tested against a set of conditions defined in the Recorder Analytics Rules. The conditions determine what actions the Recorder Analytics Rule takes. Actions include:

  • Sending a notification

  • Updating the interaction metadata

  • Placing the interaction into a specific case or folder for review by administrators or security officers responsible for case management.

Minimum score for potential target detection

The Voice Biometrics Engine server role provides a Minimum score for potential target detections threshold. This server-based threshold applies to all detection activity performed by the server role. You can adjust the threshold to control how frequently the Potentially Detected result occurs.

This server-based threshold works with the minimum detection scores configured for each Recorder Analytics Rule (Minimum Score, Minimum Score If Verified, and Minimum Score If Not Verified).

Adjusting target detection frequency

Adjust the target detection thresholds to regulate the frequency your target detection results. For example, when performing real-time detection:

Target detection result

Results from

Description

Detected

Target detection score above the Minimum Score configured in the Voiceprint Detection section of the Recorder Analytics Rule

Target detection that merits immediate action by the employee, such as transferring the interaction to the fraud team.

Adjust the Minimum Score to regulate the frequency of false positives for target detection. Your goal is to balance detection of fraudulent activity while avoiding anti-fraud measures taken for legitimate users.

Potentially Detected

Both of the following are true:

  • Target detection score below the Minimum Score configured in the Voiceprint Detection section of the Recorder Analytics Rule

  • Target detection score above the Minimum Score for potential target detection as configured for the Voice Biometrics Engine server role

Target detection may not merit immediate precautionary action by the employee but is a candidate for follow-up analysis by the fraud team.

No detection result

Target detection score below the Minimum Score for potential target detection as configured for the Voice Biometrics Engine server role

With no detection result, nothing is reported by or stored in the system for detection.

If a score is above the rule-based detection thresholds (Minimum Score, Minimum Score If Verified, and Minimum Score If Not Verified) but below the server-based Minimum score for potential target detections threshold, no detection occurs.

Employee alert for real-time detection

You can configure a Recorder Analytics Rule to provide an alert for real-time target detection.

For example, someone wants to transfer money between accounts. The person speaking on the interaction matches the voiceprint model of a target. You could immediately alert the employee of the target detection. The employee can then take more precautions, as defined by the organization, such as requiring more authentication measures before completing the transfer.

Configuring Identity Authentication and Fraud Detection

Voice Biometrics Engine role settings (Enterprise Manager Configuration and Administration Guide)