Sponsored by: FCT (PTDC/PLP/72404/2006)
Start: February 2012
End: January 2015
Project Leader: Isabel Trancoso
The goal of this project is to develop privacy-preserving frameworks for processing voice data. Processing will be performed without having access to the voice, i.e., access to any form of the speech that can be analyzed to obtain information about the talker or what they spoke. Using a combination of tools from cryptography and secure-multiparty computation we will render voice processing algorithms secure (i.e. privacy-preserving), so that the privacy of all parties is preserved. We specifically propose to develop solutions for secure speaker verification and keyword spotting problems. We envision a mechanism whereby police/security agencies could obtain publicly- accepted forms of legal sanction to look for pre-specified voices or phrases. The privacy-preserving framework will ensure that they are only notified when these occur, but will have no access to the voice data itself, thus preserving citizens’ privacy. Although this proposal addresses only speech processing techniques, the impact of this type of technique is not restricted to this area. In fact, it may affect all types of multimedia documents.