A voiceprint pre-authorization single-frame face dual-factor authentication method

Abstract

The application discloses a single-frame face double-factor authentication method of voiceprint pre-authorization. The method triggers a flow of a user offline fixed password, and first performs voiceprint verification. Only when the voiceprint passes, a camera with default constant power-off is authorized by a hardware pin of a processor to collect a single-frame face image without preview and preprocessing. Before new user collection, the user needs to perform single touch control to explicitly authorize, temporary data is only stored in a volatile running memory of the processor, features are extracted by a special hardware accelerator, the features are compared only on a server side, data zero clearing of an unshieldable hardware level is performed when the flow is terminated, and no biological feature template is stored locally, so that the method is suitable for high-security scenes such as finance.

Description

Technical Field

[0001] This invention relates to the field of biometric identification and authentication technology, specifically to a two-factor authentication method that combines voiceprint pre-authorization with single-frame face capture, suitable for high-security business scenarios such as finance. Background Technology

[0002] Existing biometric authentication technologies are mainly divided into single voiceprint authentication, single face authentication, and two-factor authentication based on video streams. Single biometric authentication suffers from drawbacks such as ease of forgery and insufficient security levels; existing two-factor authentication schemes generally rely on continuous frame acquisition from video streams, automatic camera triggering, and software-level authorization control, all of which fall within the scope of existing patent protection and pose a very high risk of infringement. Furthermore, existing schemes suffer from issues such as non-standard user authorization, temporary data residue, insufficient privacy compliance, and ambiguous technical features, failing to simultaneously meet relevant requirements. This invention addresses these shortcomings by proposing a novel hardware-level closed-loop authentication scheme. Summary of the Invention

[0003] 1. Circumvent existing solutions related to biometrics, video capture, and camera control to achieve a completely new technical solution; 2. Provides high-security two-factor authentication, with security performance superior to single biometric authentication; 3. Meets the world's most stringent privacy data protection regulations, achieving instantaneous zeroing of data throughout its entire lifecycle; 4. Construct a one-way, irreversible, hardware-locked authentication loop to prevent software cracking and side-channel attacks.

[0004] Core Innovation Points 1. Unique hardware authorization for voiceprint: The camera can only be started by the processor hardware pin level after the voiceprint verification is passed; there is no software, remote, or automatic bypass path. 2. Single-frame instantaneous acquisition: The camera operates for a short time, acquiring only one frame of raw image, with no video stream, no preview, and no preprocessing; 3. Volatile cache + hardware zeroing: Temporary data is only stored in the processor RAM, and non-maskable hardware zeroing is performed when any process terminates; 4. One-way irreversible process: The entire process can only move forward or terminate and reset, with no rollback, no retries, and no nesting; 5. Unique cloud-based comparison: There are no biological templates or feature comparison logic locally; the server only performs comparisons based on fixed thresholds. 6. Pre-authorization: A user's single touch authorization is only valid for that specific data collection and is not persistent.

[0005] System Structure 1. Offline password recognition module: Implemented locally on independent hardware, it only recognizes fixed registration / verification passwords; 2. Voiceprint Acquisition Module: Acquired via microphone; the data path is physically isolated from the password module. 3. Central Processing Unit Module: Hardware-level control of camera power supply, RAM clearing, and global reset; 4. Single-frame image acquisition module: single front-facing camera, normally powered off by default, can only be started with hardware authorization; 5. Dedicated hardware accelerator: Performs biometric extraction without software algorithm dependency; 6. Server module: Stores encrypted feature templates, performs feature comparison and one-to-one binding; 7. Independent backup power supply unit: ensures that hardware reset commands are executed first under any operating condition. Detailed Implementation

[0006] (a) New User Registration Process 1. Triggering process: The user verbally recites a fixed offline registration password, the local independent hardware controls the collection time, and the text matching threshold is met, which is the only triggering method for the process; if the password matching does not meet the threshold, the non-blockable hardware is cleared + the microphone hardware is reset + the global module is reset, the corresponding error message is displayed, and the process is terminated.

[0007] 2. Temporary voiceprint acquisition: The processor acquires voiceprints in its internal volatile RAM, which is only temporarily cached and not written to non-volatile storage. No local template is generated.

[0008] 3. Authorization Interaction: Display a fixed and unique authorization pop-up window with a countdown timer, supporting only single-time circuit breaker-style screen touch confirmation; the pop-up window clearly informs the user of the purpose of data collection, authorization validity period, and scope of data use.

[0009] 4. Authorization Denied / Timeout: If the user clicks "No" or a pop-up window times out, the backup power supply unit will perform RAM hardware clearing + full module reset + constant power off of the camera, prompting that registration has been cancelled and the process terminated.

[0010] 5. Hardware Authorization Acquisition: After the user confirms the authorization, the processor hardware pin outputs a one-time authorization level, and the single camera (default power off) starts up, completes the work in a short time, without preview, video stream, image preprocessing, or local image retention, and only performs 1 frame of raw image sampling.

[0011] 6. Feature Extraction: A dedicated hardware accelerator extracts feature vectors. If extraction fails, hardware zeroing and a full module reset are performed, indicating acquisition failure and terminating the process.

[0012] 7. Transmission and Binding: Features are encapsulated into encrypted messages and transmitted end-to-end to the server via an encrypted protocol; in case of network failure, the data is cleared and reset, a network error message is displayed, and the process terminates; the server performs feature comparison to complete a one-to-one unique binding between the user and the voiceprint-face feature.

[0013] 8. Process Closure: If binding fails, the process is reset and a registration failure message is displayed; if binding succeeds, the process is reset and a registration success message is displayed. The process terminates irreversibly in one direction.

[0014] (II) Verification Process for Existing Users 1. Triggering process: The user verbally states a fixed offline verification password, which is triggered when the matching threshold is met; if the matching fails, the password is reset, an error message is displayed, and the process terminates.

[0015] 2. Voiceprint cloud comparison: Voiceprint data is encrypted and uploaded to the server, with no local comparison logic; if the voiceprint comparison fails, it is reset, the camera is never powered on, a verification failure message is displayed, and the process is terminated.

[0016] 3. Single-frame face capture: If the voiceprint comparison meets the standard, the processor hardware is authorized to start the camera, and a single-frame raw sampling is completed in a short time. The hardware extracts features. If the extraction fails, the data is reset and a capture failure message is displayed, and the process terminates.

[0017] 4. Cloud-based face comparison: If the face comparison on the server fails, the system will reset and display a verification failure message, terminating the process. If the comparison succeeds, the system will reset and display a verification success message, granting the current valid business permissions. Permissions will automatically expire upon timeout.

[0018] 5. Global Reset: Upon termination of any process, a global hardware reset of the microphone, sensors, and RAM is performed, leaving no logs and no data residue.

[0019] Beneficial effects

[0020] 1. Avoidance: No video stream, no continuous frames, no software license, no local template, completely avoiding existing solutions; 2. High security: Voiceprint hardware pre-authorization + two-factor authentication, the security level far exceeds that of single authentication; 3. Privacy Compliance: Prior authorization, instant hardware reset, and no data retention comply with the Personal Information Protection Law of the People's Republic of China and international privacy regulations; 4. Resistance to attacks: Physically isolated from the operating system, software cracking cannot affect the core hardware logic; 5. Practicality: Applicable to general touch smart terminals, with comprehensive anomaly handling across all scenarios, and can be stably reproduced and implemented.

Claims

1. A two-factor authentication method, characterized in that, include: Users trigger the authentication process via voice commands; Voiceprint verification is performed on the user; Only after voiceprint verification is passed can a single frame of face image be captured. Two-factor comparison based on voiceprint features and single-frame face features; When the authentication process is terminated, the temporary data is cleared at the hardware level.

2. The method according to claim 1, characterized in that, The voice commands are fixed passwords that are recognized offline locally.

3. The method according to claim 1, characterized in that, The single-frame face capture process has no video stream, no image preview, and no continuous frame processing.

4. The method according to claim 1, characterized in that, The camera is in a hardware power-off state by default, and can only be activated by a hardware signal that has passed voiceprint verification.

5. The method according to claim 1, characterized in that, When a new user registers, facial data collection and feature data saving can only be performed after the user has given explicit authorization.

6. The method according to claim 5, characterized in that, When a user refuses authorization, the temporary voiceprint data is immediately cleared by hardware, and the process terminates.

7. The method according to claim 1, characterized in that, If the voiceprint verification fails, the camera will not be activated, and the process will be terminated directly.

8. The method according to claim 1, characterized in that, Feature matching is performed only on the server side; biometric templates are not stored locally.