Subaudible Tone Validation for Live Audio Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice recognition systems cannot reliably differentiate between live audio signals from a microphone and pre-recorded digital signals, making them vulnerable to hardware compromise and unauthorized access.
Innovation Solution
The system generates and outputs distinct sequences of subaudible tones, which are then received by the microphone, allowing for validation of the audio signal as live if it includes these tones, thereby authenticating the user based on the presence of these tones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voice recognition techniques are used to verify user identity, then user authentication capability is improved, but the system becomes vulnerable to hardware compromise and digital recording attacks
Solution Approach 1:
The system performs preliminary actions by outputting subaudible tones through speakers before capturing the audio signal. This preliminary emission of distinctive tones creates a temporal and spectral signature that must be present in any legitimate audio capture, preventing later spoofing attempts with pre-recorded digital signals that cannot replicate the live tone emission and capture sequence
Solution Approach 2:
Subaudible tones serve as an intermediary element between the system and the audio capture process. These tones are emitted by speakers and captured by the microphone, creating a verifiable intermediate signal that proves the audio path is live and未被 compromised. The presence of these specific subaudible tones in the captured signal acts as a mediator that validates the authenticity of the audio input
2Measurement precision
If multiple subaudible tones are output concurrently to enhance validation, then audio signal validation accuracy is improved, but device complexity increases
Solution Approach 1:
The validation process is segmented into distinct phases: first outputting a first sequence of subaudible tones, then outputting a second sequence of subaudible tones, and finally validating based on detection of these sequences. This segmentation allows the system to use multiple validation signals without overwhelming complexity, as each sequence can be independently generated and detected
Solution Approach 2:
The existing speakers and microphones are made multi-functional by using them for both normal audio operations and subaudible tone emission/detection. The speakers serve both as audio output devices and as sources of validation tones, while the microphones serve both as audio input devices and as detectors of subaudible tones. This universality avoids adding dedicated validation hardware, thereby limiting the increase in device complexity
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method effectively prevents spoofing of audio signals by ensuring that only live audio from a microphone is used for voice recognition, enhancing security by distinguishing between genuine and pre-recorded inputs.
Implementation Method 1
outputting, by a first speaker, a first sequence of subaudible tones and outputting, by a second speaker, a second sequence of subaudible tones... receiving, from a microphone, an audio signal while concurrently outputting the first and second sequences of subaudible tones
Data Source
AI summary
A method and processing system for validating an audio signal is disclosed. A valid audio signal is an audio signal that has been determined to originate from a microphone and has not been provided by a digital audio source. In some implementations, one or more speakers may generate a sequence of one or more subaudible tones. Simultaneously, a microphone near the one or more speakers captures an audio signal. If the captured audio signal includes the one or more subaudible tones from the speakers, then the audio signal is a validated.