Providing adversarial protection of speech in audio signals
AssigneeInternational Business Machines Corporation
InventorsBeat Buesser, Maria-Irina Nicolae, Ambrish Rawat, Mathieu Sinn, Ngoc Minh Tran, Martin Wistuba
Life of this patent
Estimating the term…
Abstract
Embodiments for providing adversarial protection of speech in audio signals by a processor. Security defenses on one or more audio devices may be provide against automated audio analysis of audio signals by using adversarial noise.
Claims20
A method, by one or more processors, for providing adversarial protection in audio signals, comprising:
- providing security defenses on one or more audio devices against automated audio analysis of audio signals by embedding an adversarial noise within the audio signals containing speech such that the adversarial noise distorts the speech to prevent the automated audio analysis thereof, wherein the adversarial noise is embedded to target misclassification of the speech by the automated audio analysis at one of a phoneme-level, a word-level, and a sentence-level notwithstanding the speech is maintained to be intelligible by a user.
A system for providing adversarial protection in audio signals, comprising:
- one or more computers with executable instructions that when executed cause the system to:
- provide security defenses on one or more audio devices against automated audio analysis of audio signals by embedding an adversarial noise within the audio signals containing speech such that the adversarial noise distorts the speech to prevent the automated audio analysis thereof, wherein the adversarial noise is embedded to target misclassification of the speech by the automated audio analysis at one of a phoneme-level, a word-level, and a sentence-level notwithstanding the speech is maintained to be intelligible by a user.
A computer program product, for providing adversarial protection in audio signals by one or more processors, the computer program product comprising a non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions comprising:
- an executable portion that provides security defenses on one or more audio devices against automated audio analysis of audio signals by embedding an adversarial noise within the audio signals containing speech such that the adversarial noise distorts the speech to prevent the automated audio analysis thereof, wherein the adversarial noise is embedded to target misclassification of the speech by the automated audio analysis at one of a phoneme-level, a word-level, and a sentence-level notwithstanding the speech is maintained to be intelligible by a user.
Dependencies are inferred from the claim text.