Call Audio Deepfake Detection Using Voice Liveness Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic speaker verification systems are vulnerable to deepfake attacks, where malicious actors use machine-generated speech to bypass authentication, posing a threat to secure systems and information access.
Innovation Solution
Implement a system that combines speech-phrase repetition detection, background change detection, and passive voice-liveness detection, using machine-learning architectures for deepfake detection, and adapts to new attacks through automated model updates based on human feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional automatic speaker verification systems are used for authentication, then authentication speed and efficiency are improved, but the system becomes vulnerable to deepfake attacks and presentation attacks
Solution Approach 1:
The authentication system is divided into multiple independent detection modules: speech-phrase repetition detection, background change detection, and passive voice-liveness detection. Each module analyzes different aspects of the audio signal independently, and their results are combined to make the final authentication decision. This segmentation allows the system to maintain high authentication speed while improving security by checking multiple hypotheses simultaneously.
Solution Approach 2:
The patent combines three different detection methods (speech-phrase repetition detection, background change detection, and passive voice-liveness detection) into a unified authentication system. These methods are merged through a decision-making framework that integrates their results, creating a comprehensive deepfake detection system that maintains efficiency while improving reliability through multiple detection angles.
2Measurement precision
If multiple detection methods are combined for deepfake detection, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The detection system is segmented into three distinct modules, each handling a specific aspect of deepfake detection: speech-phrase repetition detection (analyzing temporal patterns), background change detection (monitoring environmental audio), and passive voice-liveness detection (assessing natural speech characteristics). This segmentation reduces overall complexity by making each module focused and manageable while improving accuracy through specialized detection.
Solution Approach 2:
The system employs a universal decision-making framework that can integrate multiple detection methods and adapt to different attack types. The framework is designed to handle various presentation attacks (replay attacks, deepfake attacks, and other spoofing attempts) using the same architectural structure, making the system versatile and reducing complexity through reuse of common processing components.
3Adaptability or versatility
If the detection system uses automated model updates based on human feedback, then adaptability to new attack types is improved, but processing time and computational resources increase
Solution Approach 1:
The system implements a feedback mechanism where human feedback on detection results is collected and used to automatically update the detection models. This feedback loop allows the system to adapt to new presentation attack techniques by learning from real-world encounters. The feedback mechanism is designed to be efficient, using only necessary computational resources for model updates while maintaining fast real-time detection during normal operation.
Data Source
AI summary
Disclosed are systems and methods including software processes executed by a server that detect audio-based synthetic speech (“deepfakes”) in a call conversation. The server applies an NLP engine to transcribe call audio and analyze the text for anomalous patterns to detect synthetic speech. Additionally or alternatively, the server executes a voice “liveness” detection system for detecting machine speech, such as synthetic speech or replayed speech. The system performs phrase repetition detection, background change detection, and passive voice liveness detection in call audio signals to detect liveness of a speech utterance. An automated model update module allows the liveness detection model to adapt to new types of presentation attacks, based on the human provided feedback.