Live Speech Detection Certificate for Replay Attack Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition systems are vulnerable to replay attacks and false activations due to the inability to reliably distinguish live speech from recorded speech, leading to unintended actions by devices.
Innovation Solution
A method and device that perform live speech detection, generate a certificate by encrypting the detection output, and transmit it to a separate device for authentication, ensuring secure verification of live speech before processing commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speaker recognition systems use voice biometric comparison with enrolled models, then they can verify speaker identity, but they become vulnerable to replay attacks using recorded speech
Solution Approach 1:
The system performs live speech detection before processing the speech command. A certificate is generated by encrypting the live speech detection output and transmitted to the second device. This preliminary verification step ensures that only live speech is processed, preventing replay attacks before they can compromise the system.
Solution Approach 2:
A certificate serving as an intermediary element is introduced between the speech signal and the authentication process. The certificate contains encrypted live speech detection output and is transmitted from the first device to the second device, mediating the verification process and providing evidence that the speech is live without revealing the detection methodology.
2Reliability
If the system transmits speech signals to a separate device for authentication, then security is improved, but device complexity increases
Solution Approach 1:
The authentication system is segmented into two separate devices: a first device that captures speech and performs live speech detection, and a second device that receives the speech signal and certificate for authentication. This segmentation distributes complexity across multiple devices, allowing each to be optimized for its specific function while maintaining overall system security.
Data Source
AI summary
A method of authenticating a speech signal in a first device comprises receiving a speech signal, and performing a live speech detection process to determine whether the received signal represents live speech. The live speech detection process generates a live speech detection output. A certificate is formed by encrypting at least the live speech detection output. The received signal, and the certificate, are transmitted to a separate second device.