Skip to main content

Secure Access Front-End

Objective

The main objective of the project is to develop a product prototype that will increase the services access security with no increase of the service complexity. The prototype will check the identity of the user analysing the speaker's voice by means of speech recognition and speaker verification techniques and will work both in the mobile and the IP networks. The system will initially work in six European Languages (Spanish, English, French, Catalan, Galician and Basque) even though the system architecture will easily allow the addition of new languages. By reaching these objectives, the SAFE project contributes to improving the accessibility and quality of public services and will help to develop a multilingual information society, ensuring secure universally available access to special services like electronic banking, private vocal messaging, commercial transactions, etc.

Objectives:
The result of this project is a multilingual secure access front-end that will check the identity of the user of a service through the mobile or the IP network (G.723, G.729). The system prototype will be based on speech recognition and speaker verification techniques and will use a decision mechanism to combine the outputs of both modules. The main objective of the project is to develop a product prototype that will increase the services access security with no increase of the service complexity. The system will initially work in six European Languages (Spanish, English, French, Catalan, Galician and Basque) even though the system architecture will easily allow the addition of new languages. By reaching these objectives, the SAFE project contributes to improving the accessibility and quality of public services and will help to develop a multilingual information society, ensuring secure universally available access to special services.

Work description:
The SAFE project will develop a secure access front-end prototype. This prototype will combine speech recognition and speaker verification techniques to recognise the user Personal Identification Number (PIN) and simultaneously check the user identity on the speech produced by (him) herself to pronounce the PIN. The prototype will be built by using the already available software libraries of the Telefonica' s Speech Software (TSS) and the new speaker verification modules that will be integrated in this project. The SAFE project directly addresses one of the goals of the TRIAL projects, which is the adaptation of already available leading edge technology to develop new products for service applications. We will first adapt and test the current speaker verification technology to the mobile and IP (G.723, G.729) networks and test several already implemented methods for dealing with the multilingual speaker verification task, i.e.: there are several ways of training the Universal Background Model (UBM) that we will use for detecting and rejecting impostors and we will select the one that behaves the best in the multilingual case. These tasks will be carried out using already existing speech databases and a specific multi-session database that will be recorded in the project. This multi-session database will be used to check the performance of the system in the mobile and IP networks for time-spaced utterances from the same speakers. Also, the speech recognition product of TID will be adapted to the IP network and then tested. Later on the speaker verification system will be integrated into the Telefonica' s Speech Software product. Finally, SAFE will be implemented and tested with real users. Along the course of the evaluation we will publish in relevant international conferences and/or magazines the main results of the project.

Milestones:
The SAFE project is composed of 16 milestones that correspond to the completion of 16 project deliverables. The expected result is a secure access front-end prototype, which could be used in any already existing service or in new services requiring a high level of access security with no increase in the system complexity to the user.

Funding Scheme

ACM - Preparatory, accompanying and support measures
Leaflet | Map data © OpenStreetMap contributors, Credit: EC-GISCO, © EuroGeographics for the administrative boundaries

Coordinator

TELEFONICA INVESTIGACION Y DESARROLLO SA
Address
Calle Emilio Vargas 6
28043 Madrid
Spain

Participants (2)

INFORMATICA EL CORTE INGLES, S.A.
Spain
Address
Travesia Costa Brava 4
28034 Madrid
UNIVERSIDAD POLITECNICA DE MADRID
Spain
Address
Avenida Ramiro De Maeztu 7
28040 Madrid