Acoustic Location Authentication Without Exposing Speech Content
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Healthcare providers face authentication burdens when rapidly moving between computing devices and require privacy protection for audio recordings used in authentication processes.
Innovation Solution
A device and method for privacy-preserving authentication using trained models to generate privacy-preserving acoustic representations of audio segments from different microphones, determining similarity based on these representations to authenticate devices as being associated with a similar location, thereby allowing secure authentication without transmitting sensitive speech content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional authentication methods are used at each computing device, then security is maintained, but healthcare providers face significant authentication burden and time loss when moving between devices
Solution Approach 1:
The system performs preliminary authentication by capturing audio data from the first computing device and generating acoustic representations before the healthcare provider moves to the second device. This pre-authentication process eliminates the need for repeated authentication at each device, resolving the contradiction between security and authentication convenience.
Solution Approach 2:
The system creates acoustic representations (copies) of the healthcare provider's voice from the first device and uses these copies to authenticate at the second device. This copying mechanism allows authentication without requiring the provider to physically present themselves or undergo time-consuming verification procedures at each device.
2Measurement precision
If audio data is transmitted for authentication purposes, then authentication accuracy is improved, but privacy of audio content is compromised
Solution Approach 1:
The system extracts only the essential acoustic features (pitch, frequency, timbre, formants) from the audio data while leaving out the actual speech content. This extraction process maintains authentication accuracy by preserving voice characteristics while protecting privacy by removing the meaningful speech information that would compromise audio privacy.
Solution Approach 2:
The system applies different processing qualities to different aspects of the audio data: high-fidelity processing is applied to acoustic features needed for authentication, while the speech content is degraded or removed. This local quality differentiation allows the system to maintain authentication accuracy where needed while protecting privacy where not needed.
Data Source
AI summary
A computer-implemented method for privacy-preserving authentication includes: receiving a first privacy-preserving acoustic representation of a first audio segment associated with a first microphone from a first computing device, the first privacy-preserving acoustic representation being locally generated on the first computing device by a trained model based on the first audio segment; receiving a second privacy-preserving acoustic representation of a second audio segment associated with a second microphone from a second computing device, the second audio segment being generated by the second microphone contemporaneously with the first audio segment; generating a similarity metric score based on the first privacy-preserving acoustic representation and the second privacy-preserving acoustic representation; determining that the first microphone and the second microphone are associated with a substantially similar location based on the similarity metric score; and authenticating that the first microphone and the second microphone are associated with the substantially similar location.