Background

The Conversational Biometric Authentication in Real Time (C-BART) system is an implementation of conversational biometrics in the telephony environment which allows for user identification and verification using spoken natural language. Conversational biometrics combines two sources for authentication: 1) physical speech biometrics (the voice-print) and 2) user knowledge (such as passwords and personal information). The combination of the two information sources increases the security and reliability and provides a flexible framework for various authentication scenarios so as to maximize user convenience.

Technologies utilized to enable conversational biometrics include acoustic speaker recognition, speech recognition, natural language understanding and dialog management. In our prototype, the verification consists of one or several short interviews involving randomly asked authentication questions and an acoustic voice-print check. The length of a session depends on the correctness of the answers and the estimated voice-print confidence. Users can enroll into the system via an HTML-form and a telephone server. Measured on real data, for a 3% false rejection, casual impostors are falsely accepted by the system in less than 0.00001% of cases.

Demonstration Recordings

Please save and then run the content locally if the audio and video links below do not automatically play when selected.

Scenario 1: An authentic user accesses his account An authorized user is granted bank access privileges through the use of the C-BART system. The C-BART system authenticates the customer via an acoustic and knowledge profile analysis. Due to an appropriate voice match only one question is asked to successfully verify the caller. (The user is identified as "Ran".) | Download the 53 second AUDIO (448kB WAV) or VIDEO (1072kB WMV) recordings.

Scenario 2: An impostor is attempting to compromise a user's account. Besides the natural voice mismatch, the impostor has no knowledge available to answer the biometrics questions correctly. | Download the 32 second AUDIO (288kB WAV) or VIDEO (848kB WMV) recordings.

Scenario 3: A (well informed) impostor is trying to compromise a user's account. In this scenario, the impostor is in possession of the complete list of correct answers to the biometrics questions of the true user. In practice, capturing the complete question and answer pool (for example by eaves-dropping) is made difficult by randomizing the interviews and asking different questions in consecutive sessions. | Download the 43 second AUDIO (368kB WAV) or VIDEO (1184kB WMV) recordings.

Publications

• Ganesh N. Ramaswamy, Jirí Navrátil, Upendra V. Chaudhari and Ran D. Zilca, "The IBM System for the NIST-2002 Cellular Speaker Verification Evaluation", Proc. of the International Conference on Acoustics, Speech and Signal Processing (ICASSP), Vol 2, Pages 61-64, 2003.

• Ganesh N. Ramaswamy, Ran D. Zilca and Oleg Alecksandrovich, "A Programmable Policy Manager for Conversational Biometrics", Proc. of Eurospeech, Vol 3, Pages 1957-1960, 2003.

• Stéphane H. Maes, Jirí Navrátil and Upendra V. Chaudhari, "Conversational Speech Biometrics", Chapter in E-Commerce Agents Marketplace Solutions, Security Issues, and Supply and Demand, J. Liu and Y. Ye (Eds.):  Springer Verlag, Pages 166-179, 2001.

• Jirí Navrátil, Jan Kleindienst and Stéphane H. Maes, "An Instantiable Speech Biometrics Module with Natural Language Interface: Implementation in the Telephony Environment", Proc. of the International Conference on Acoustics, Speech and Signal Processing (ICASSP), Istanbul, Turkey, June 2000.