Cybersecurity for sensitive-information utterances in interactive voice sessions using risk profiles

A computing system comprising:

• a memory configured to store first audio data representing one or more initial utterances of a user during an interactive voice session with an interactive voice system, wherein a risk profile of the interactive voice system comprises a risk score for each class of sensitive information in a plurality of classes of sensitive information;
• and
• processing circuitry configured to:
• for each respective class of sensitive information in the classes of sensitive information:
• generate, based on the first audio data, a confidence score for the respective class of sensitive information that indicates a level of confidence that a subsequent utterance of the user in the interactive voice session will contain sensitive information in the respective class of sensitive information, wherein the subsequent utterance follows the one or more initial utterances in time;
• determine a combined score for the respective class of sensitive information based on the risk score for the respective class of sensitive information and the confidence score for the respective class of sensitive information;
• and
• obtain second audio data representing the subsequent utterance;
• determine, based on the combined scores for the classes of sensitive information, whether to transmit the second audio data to the interactive voice system;
• and
• based on a determination to transmit the second audio data to the interactive voice system, transmit the second audio data to the interactive voice system.

A non-transitory computer-readable storage medium comprising instructions that, when executed cause processing circuitry of a computing system to:

• obtain first audio data representing one or more initial utterances of a user during an interactive voice session with an interactive voice system, wherein a risk profile of the interactive voice system comprises a risk score for each class of sensitive information in a plurality of classes of sensitive information;
• for each respective class of sensitive information in the classes of sensitive information:
• generate, based on the first audio data, a confidence score for the respective class of sensitive information that indicates a level of confidence that a subsequent utterance of the user in the interactive voice session will contain sensitive information in the respective class of sensitive information, wherein the subsequent utterance follows the one or more initial utterances in time;
• and
• determine a combined score for the respective class of sensitive information based on the risk score for the respective class of sensitive information and the confidence score for the respective class of sensitive information;
• obtain second audio data representing the subsequent utterance;
• determine, for each respective class of sensitive information, a combined score based on the respective confidence score and the respective risk score;
• determine, based on the combined scores for the classes of sensitive information, whether to transmit the second audio data to the interactive voice system;
• and
• based on a determination to transmit the second audio data to the interactive voice system, transmit the second audio data to the interactive voice system.