Voice Biometric Authentication Against Synthetic Speech Spoofing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice biometric systems are susceptible to synthetic voice attacks due to their inability to adapt to changes in a person's voice over time, leading to increased false rejection rates and decreased security, which can compromise the integrity of the system and expose users to fraud.
Innovation Solution
A method and system that analyze voice biometric data in real-time, updating the data with captured voice samples, modifying it to include background noise and altering frequency modulation, and securely storing it using blockchain technology to prevent unauthorized alterations, while employing adaptive thresholds and anomaly detection algorithms to distinguish between genuine and synthetic speech.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If voice biometric data is stored statically without updates, then system simplicity is maintained, but the system becomes susceptible to synthetic voice attacks and cannot adapt to voice changes over time
Solution Approach 1:
The patent implements dynamic updating of voice biometric data records by capturing new voice samples during authentication transactions and incrementally updating the stored template. This allows the system to adapt to voice changes over time while maintaining a manageable update frequency (e.g., after a threshold number of transactions or time-based intervals), thus resolving the contradiction between adaptability and complexity.
Solution Approach 2:
The system modifies parameters of the stored voice biometric data by blending new voice samples with existing templates using weighted averages or other combination techniques. This parameter transformation allows gradual adaptation to voice changes while maintaining system simplicity through mathematical operations rather than complex restructuring.
2Adaptability or versatility
If voice biometric data is updated frequently with new samples, then adaptability to voice changes improves, but computational resources and processing time increase
Solution Approach 1:
The patent implements periodic updating mechanisms where voice biometric data is updated after a threshold number of authentication transactions or after a specified time interval has elapsed. This periodic approach balances adaptability with processing efficiency by avoiding continuous updates while still capturing significant voice changes over time.
Solution Approach 2:
The system performs partial updates by blending new voice samples with existing templates rather than complete re-enrollment, using techniques like weighted averaging where the new sample contributes a fraction (e.g., 10-20%) to the updated template. This partial action reduces processing requirements while maintaining adaptability.
3Reliability
If security measures are enhanced with anomaly detection and blockchain technology, then protection against synthetic voice attacks improves, but system complexity and computational overhead increase
Solution Approach 1:
The patent introduces intermediary components such as anomaly detection algorithms that analyze voice characteristics for signs of synthesis, and blockchain technology that provides tamper-proof recording of authentication transactions. These intermediaries enhance security without requiring complete system redesign, allowing incremental integration of security measures.
Solution Approach 2:
The system performs preliminary anomaly detection on incoming voice samples before processing authentication, checking for characteristics indicative of synthetic speech. This preliminary action filters potential threats early in the process, preventing wasted computational resources on fraudulent attempts while maintaining security.
4Measurement precision
If anomaly detection algorithms are continuously updated to match synthetic voice evolution, then detection accuracy improves, but computational resources and algorithm complexity increase
Solution Approach 1:
The patent implements feedback mechanisms where detection results from authentication transactions are used to continuously refine and update anomaly detection algorithms. Successful detections and missed anomalies provide training data that improves future detection accuracy without requiring constant high-computational resource allocation, as updates can occur during low-utilization periods.
Solution Approach 2:
The system employs lightweight, computationally efficient anomaly detection algorithms that can be quickly executed during authentication transactions, rather than relying on heavy, resource-intensive models. These simpler algorithms consume fewer computational resources while maintaining adequate detection accuracy for the specific threat landscape.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
A method for enhancing the security of voice biometric data is provided that includes receiving, by an electronic device, voice biometric data of a user captured while the user was speaking and analyzing the context in which the received voice biometric data was captured. The context includes environmental and situational factors. The method also includes analyzing characteristics of the received voice biometric data to detect anomalies associated with synthetic speech, and conducting a biometric authentication transaction based on the voice biometric data in response to determining the received voice biometric data is free of anomalies. In response to successfully authenticating the user, the method includes updating record voice biometric data of the user with the received voice biometric data. Moreover, the method includes modifying the updated record voice biometric data and securely storing the modified updated record voice biometric data.