Voice Authentication for Secure Banking

Contactless authentication using voice recognition to improve security, convenience, and cross-platform compatibility compared to conventional authentication methods that require contact. The authentication is performed by converting an audio sample from a user device into a processed format and transmitting it to a speech recognition module. The speech recognition module compares the processed audio against user voice models to determine if it matches. If the match is successful, the user is authenticated. This allows contactless voice-based authentication that can be used on any device with a microphone and speaker, without requiring dedicated authentication devices or manual entry.

Source

Secure electronic financial transactions system that uses AI, biometrics, and multimodal data analysis to prevent fraud and improve user experience. The system collects multimodal data from visual, auditory, and tactile sensors during transactions. It uses AI modules like biometric authentication, transaction anomaly detection, geospatial analysis, behavioral analysis, and third-party data integration to analyze this data for fraud prevention. The system also communicates with users through modalities like vision, audio, touch, taste, smell, temperature, pain, and balance to address fraud concerns or request verification.

Source

Contactless authentication system using voice recognition that allows user devices to authenticate without physical contact or dedicated authentication devices. It involves receiving user voice data, converting it to a standard format, sending it to a remote server for voice matching, and using the match result to authenticate the user. This allows cross-platform compatibility since the authentication is performed remotely and doesn't require device-specific formats.

Source

Quickly realizing highly secure voice authentication without requiring learning a global voiceprint model, using a calculation method like the Viterbi algorithm, or calculating an average reliability value. The method involves extracting a voice feature vector from a user's registered voice, generating a temporary password, presenting it to the user, receiving their voice reading the password, and authenticating using both voice feature vectors and the password. This allows fast voice authentication without complex speech recognition processing.

Source

Privacy-protecting voiceprint authentication to prevent exposure of user biometric data in voice authentication systems. The method involves synthesizing a voiceprint template by combining extracted voice features from a user's voice spoken in multiple modes, rather than sending the actual voice. This synthesized template is used for authentication instead of the original voice. This prevents exposing the user's actual voice during registration and authentication. The synthesized voiceprint template is sent to the server for storage, which cannot be reversed to obtain the user's original voice.

Source

Enhancing user authentication using a smart device like a smart speaker to improve security and reliability. The method involves using transaction-based authentication questions like "What was your credit card purchase on May 12?" instead of static questions. The user's voice is analyzed during the responses to determine if it matches an expected voiceprint. If so, fewer questions are needed for authentication. If not, more questions are required. This combines voice analysis with dynamic transaction-based questions to prevent spoofing of prerecorded voices.

Source

Voice verification method and network equipment that enhances the accuracy and security of voice password verification while also providing real-time performance. The method involves using a voice recognition server to compare a user's live voice input against a stored voice print, instead of using pre-recorded voice files. The live comparison reduces the risk of stolen usernames and forgotten passwords. The voice recognition server extracts speech features from the live input and compares them against the stored voice print in real-time. This avoids issues with voice details changing due to emotion or state differences between password setting and verification.

Source

Multi-factor authentication system that includes voice verification as a factor. The system allows users to enroll their voice for authentication by providing a few sample utterances. It trains a text-independent voice model using the samples to capture distinctive aural characteristics of the user's voice. During authentication, the system compares the user's live voice to the trained model to verify their identity. This provides an additional factor of authentication beyond passwords that is more convenient than using hardware tokens or biometrics.

Source

Exchanging data for secure transactions over voice channels to enable voice-based payment and other transactions. The method involves monitoring voice conversations for transaction requests, extracting details from the voice, confirming with the user, and carrying out the transaction using the extracted data. It uses voice encoding to embed transaction requests in audio and voice decoding to extract them. This allows performing transactions without separate channels or devices.

Source

Continuously authenticating a person by voice to extract their voice from a mixed audio stream or continuously monitor if an ongoing call is with a specific person. The technique involves continuously performing voiceprint authentication on the acquired voice in time series using stored voiceprint data. Authentication result values representing estimation degrees are generated. Thresholds are used to selectively pass or discard the voice based on how closely it matches the target. This allows extracting just the target person's voice from a mixed stream or continuously monitoring if a call is with the target.

Source

Enhanced security for accessing resources like bank accounts using biometric verification at the time of transaction in addition to PIN entry. The method involves comparing current biometric data with historical data to ensure consistency. This prevents unauthorized use of stolen cards or compromised PINs. A generative adversarial network (GAN) is trained to enhance biometric data and generate distributions for verification. It analyzes discrepancies and determines if multiple users could match the data.

Source

Enabling voice biometric authentication for user accounts that allows non-intrusive collection of voice data during normal calls to enroll users and authenticate them without disrupting the call flow. The system enhances audio quality during calls to capture voice samples for enrollment. It filters background noise to focus on high-definition voice characteristics. This minimally intrusive enrollment is performed during normal conversations between users and agents. The enhanced audio is stored and converted to voiceprints for authentication.

Source

Authenticating user identity in digital services and voice calls using voice analysis instead of passwords or security questions. The method involves comparing a user's recorded voice speaking a unique identifier against their stored voice record to authenticate. Speech recognition extracts the identifier, and voice comparison verifies identity. It enables hands-free login and call access without manual entry or remembering passwords.

Source

Training a machine learning model for simultaneously authenticating a user and determining liveness of the user's biometric voice data. The training involves using a dataset of audio signals with user identities and flags indicating if the audio was live or synthetic. The model learns embeddings from the audio signals and optimizes losses like triplet loss to successfully verify user identity and liveness. This joint training allows the model to authenticate users and detect spoofing attacks like voice cloning or conversion.

Source

A terminal device that allows user authentication using multiple modalities to reduce burden and improve success rates. The device can authenticate using face recognition or voice recognition. If face authentication fails, it switches to voice authentication using vocal tract characteristics. This allows fallback authentication if one modality fails.

Source

Enhancing online account security by adding voice biometric authentication to the two-factor authentication process. When a user tries to log in to an online account, in addition to entering a password, they are prompted to provide a voice sample. The system compares the recorded voice to a known voice biometric profile for that user. If the match is successful, the user is granted access to the account. This helps prevent unauthorized access using stolen passwords or compromised SIM cards, as it requires the user's live voice to authenticate.

Source

Authenticating users via voice prints in self-service systems to prevent voice phishing attacks. The method involves generating a text-independent voice print for user enrollment. When a user requests authentication, a passphrase is selected based on comparing it against text-dependent voice biometric models. The user is asked to repeat the selected passphrase. Their recorded response is compared against the text-independent voice print to authenticate them. This prevents phishing attacks as the passphrase is unique per session.

Source

Securely validating credentials like bank account numbers without exposing the actual credentials. The method involves training a neural network to infer validity of encrypted credentials. The neural network is then encrypted using the same homomorphic encryption as the credentials. This allows validating encrypted credentials without decrypting them. The encrypted validity indicators can be provided back to the original source without exposing the original encrypted credentials.

Source

Self-service correction terminal anti-counterfeiting identity verification method that uses dynamic speech with stop commands and image cues to prevent forgery. The method involves generating anti-counterfeiting rules with dynamic sentences and stop commands for voice authentication. These rules are displayed to users. They read the sentences with stops, and the terminal collects the voice. It extracts voiceprint features, recognizes speech segments, and compares against the dynamic rules. This makes it harder to forge voices since stops and images are required.

Source

Mobile banking application security authentication using voice recognition instead of passwords or biometrics. The method involves comparing voice recordings to stored voiceprints to authenticate users. When a user requests a banking action, the app plays a prompt and records their voice. It compares the recorded voice to the user's stored voiceprint to verify identity. If no voiceprint exists, it prompts the user to record one. This provides convenience without requiring active participation like fingerprints or passwords.

Source