Adaptiivne kõnepõhine juhuväljakutsete tugi biomeetrilisele autentimisele
Date
2012
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Tartu Ülikool
Abstract
Käesoleva bakalaureusetöö eesmärgiks oli arendada välja kõnetuvastusprogramm, mida saaks kasutada vokaalsete juhuväljakutse tarvis. Programmi eesmärgiks oli anda üks võimalik lahendus kõnepõhilise biomeetrilise autentimise kesksele turvaprobleemile – taasesitusrünnetele. Programm põhineb vabavaralisel PocketSphinxi kõnetuvastuse tööriistal ning on kirjutatud Pythoni programmeerimiskeeles.
Loodud rakendus koosneb kahest osast: kasutajaliidesega varustatud demonstratsiooniprogrammist ja käsurea utiilidist. Kasutajaliidesega rakendus sobib kõnetuvastusteegi võimete demonstreerimiseks, käsurea utiliiti saab aga kasutada mis tahes teisele programmile kõnetuvastusvõimekuse lisamiseks.
Kasutajaliidesega rakenduses saab kasutaja oma hääle abil programmiga vahetult suheldes avada näitlikustamiseks loodud demoprogrammi ust. Kasutaja peab ütlema õige numbrite jada või pildile vastava sõna inglise keeles, et programmi poolt autoriseeritud saada.
Mõlemat loodud rakendust saab seadistada luues oma keelemudeleid või muutes demorakenduse puhul numbriliste juhuväljakutsete pikkust.
The aim of this thesis was to develop a speech recognition application which could be used for vocal random challenges. The goal of the application was to provide a solution to the central problem for voice-based biometric authentication – replay attacks. This piece of software is based on the PocketSphinx speech recognition toolkit and is written in the Python programming language. The resulting application is composed of two parts: a demonstration application with a GUI interface, and a command line utility. The GUI application is suitable for demonstrating the capabilities of the speech recognition toolkit, whereas the command line utility can be used to add speech recognition capabilities to virtually any application. The user can interact with the door of the GUI application by using his or her voice. The user must utter the correct word corresponding to the picture in English or say the sequence of digits in order to be authenticated. Both of the applications can be configured by generating language models, or by changing the length of the random challenges for the demonstration application.
The aim of this thesis was to develop a speech recognition application which could be used for vocal random challenges. The goal of the application was to provide a solution to the central problem for voice-based biometric authentication – replay attacks. This piece of software is based on the PocketSphinx speech recognition toolkit and is written in the Python programming language. The resulting application is composed of two parts: a demonstration application with a GUI interface, and a command line utility. The GUI application is suitable for demonstrating the capabilities of the speech recognition toolkit, whereas the command line utility can be used to add speech recognition capabilities to virtually any application. The user can interact with the door of the GUI application by using his or her voice. The user must utter the correct word corresponding to the picture in English or say the sequence of digits in order to be authenticated. Both of the applications can be configured by generating language models, or by changing the length of the random challenges for the demonstration application.