Disclosed are systems and methods including computing-processes executing machine-learning architectures extract vectors representing disparate types of data and output predicted identities of users accessing computing services, without express identity assertions, and across multiple computing services, analyzing data from multiple modalities, for various user devices, and agnostic to architectures hosting the disparate computing service. The system invokes the identification operations of the machine-learning architecture, which extracts biometric embeddings from biometric data and context embeddings representing all or most of the types of metadata features analyzed by the system. The context embeddings help identify a subset of potentially matching identities of possible users, which limits the number of biometric-prints the system compares against an inbound biometric embedding for authentication. The types of extracted features originate from multiple modalities, including metadata from data communications, audio signals, and images. In this way, the embodiments apply a multi-modality machine-learning architecture.

Disclosed are systems and methods including computing-processes executing machine-learning architectures extract vectors representing disparate types of data and output predicted identities of users accessing computing services, without express identity assertions, and across multiple computing services, analyzing data from multiple modalities, for various user devices, and agnostic to architectures hosting the disparate computing service. The system invokes the identification operations of the machine-learning architecture, which extracts biometric embeddings from biometric data and context embeddings representing all or most of the types of metadata features analyzed by the system. The context embeddings help identify a subset of potentially matching identities of possible users, which limits the number of biometric-prints the system compares against an inbound biometric embedding for authentication. The types of extracted features originate from multiple modalities, including metadata from data communications, audio signals, and images. In this way, the embodiments apply a multi-modality machine-learning architecture.

A system is provided for real-time fraud detection with a fraudster voice print watchlist of repetitive fraudster phrases. The system includes a processor and a computer readable medium operably coupled thereto, to perform fraud prevention operations which include detecting a voice communication session having an audio signal of a user, accessing the fraudster voice print watchlist comprising a plurality of fraudster voice prints of the repetitive fraudster phrases, generating a voice print of the user using the audio signal, monitoring the user for real-time fraud detection using the fraudster voice print watchlist and the voice print of the user, and determining, based on the monitoring, whether the voice print of the user meets or exceeds a scoring threshold for matching with one or more of the plurality of fraudster voice prints from the fraudster voice print watchlist during the voice communication session.

A system is provided for real-time fraud detection with a fraudster voice print watchlist of repetitive fraudster phrases. The system includes a processor and a computer readable medium operably coupled thereto, to perform fraud prevention operations which include detecting a voice communication session having an audio signal of a user, accessing the fraudster voice print watchlist comprising a plurality of fraudster voice prints of the repetitive fraudster phrases, generating a voice print of the user using the audio signal, monitoring the user for real-time fraud detection using the fraudster voice print watchlist and the voice print of the user, and determining, based on the monitoring, whether the voice print of the user meets or exceeds a scoring threshold for matching with one or more of the plurality of fraudster voice prints from the fraudster voice print watchlist during the voice communication session.

Various embodiments of the disclosure provide a method and apparatus for processing a voice command in an electronic device. According to various embodiments of the disclosure, the electronic device includes a microphone, a memory, and a processor operatively coupled to the microphone and the memory. The processor may be configured to wake up on the basis of detection of a voice call command, calculate a score related to recognition of the voice call command, share the score with an external device, decide whether to execute a voice command on the basis of the score, and process the voice command on the basis of the decision result. Various embodiments are possible.

Various embodiments of the disclosure provide a method and apparatus for processing a voice command in an electronic device. According to various embodiments of the disclosure, the electronic device includes a microphone, a memory, and a processor operatively coupled to the microphone and the memory. The processor may be configured to wake up on the basis of detection of a voice call command, calculate a score related to recognition of the voice call command, share the score with an external device, decide whether to execute a voice command on the basis of the score, and process the voice command on the basis of the decision result. Various embodiments are possible.

A system is provided for fraud prevention upscaling with a fraudster voice print watchlist. The system includes a processor and a computer readable medium operably coupled thereto, to perform fraud prevention operations which include receiving a first voice print of a user during a voice authentication request, accessing the fraudster voice print watchlist comprising voice print representatives for a plurality of voice print clusters each having one or more of a plurality of voice prints identified as fraudulent for a voice biometric system, determining that one or more of the voice print representatives in the fraudster voice print watchlist meets or exceeds a first biometric threshold for risk detection of the first voice print during the fraud prevention operations, and determining whether the first voice print matches a first one of the plurality of voice print clusters.

A system is provided for fraud prevention upscaling with a fraudster voice print watchlist. The system includes a processor and a computer readable medium operably coupled thereto, to perform fraud prevention operations which include receiving a first voice print of a user during a voice authentication request, accessing the fraudster voice print watchlist comprising voice print representatives for a plurality of voice print clusters each having one or more of a plurality of voice prints identified as fraudulent for a voice biometric system, determining that one or more of the voice print representatives in the fraudster voice print watchlist meets or exceeds a first biometric threshold for risk detection of the first voice print during the fraud prevention operations, and determining whether the first voice print matches a first one of the plurality of voice print clusters.

Techniques are described for user account matching based on natural language utterances. In an example, a computer system receives a set of words, a voice print, and offer data about an offer based at least in part on a natural language utterance at a user device. The computer system determines a set of user accounts based at least in part on the set of words and determines, from this set, a first user account based at least in part on the voice print. The first user account is associated with a first user identifier. The computer system determines that the offer is associated with a second user account that is further associated with a second user identifier. The computer system generates associations of the user accounts with user identifiers and with the offer.

Techniques are described for user account matching based on natural language utterances. In an example, a computer system receives a set of words, a voice print, and offer data about an offer based at least in part on a natural language utterance at a user device. The computer system determines a set of user accounts based at least in part on the set of words and determines, from this set, a first user account based at least in part on the voice print. The first user account is associated with a first user identifier. The computer system determines that the offer is associated with a second user account that is further associated with a second user identifier. The computer system generates associations of the user accounts with user identifiers and with the offer.