Privacy Preserving Speaker Verification using Secure Binary Embeddings


Revision as of 22:32, 2 October 2013 by Rdmr (Talk | contribs) (Created page with "__NOTOC__ {{infobox|name=José Portêlo |username=portelo |contact=portelo |phone=+351-213-100-514 |fax=+351-213-145-843 }} == Date == * 15:00, Friday, October 4<sup>th</sup>, ...")

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
José Portêlo


  • 15:00, Friday, October 4th, 2013
  • Room 336



Conventional remote speaker verification systems rely on the system to have access to the user’s recordings, or features derived from them, and also a model of the user’s voice. In the proposed approach, the system has access to none of them. The features extracted from the user’s recordings are transformed to bit strings in a way that allows the computation of approximate distances, instead of exact ones. The key to the transformation uses a hashing scheme known as Secure Binary Embeddings. An SVM classifier with a modified kernel operates on the hashes. This allows speaker verification to be performed without exposing speaker data. Experiments showed that the secure system yielded similar results as its non-private counterpart. This approach may be extended to other types of biometric authentication.