Decrease size
Reset to Default
Increase size

Development of anti-spoofing measure for speech interfaces to prevent computer generated voice hacking

Primary Information

Domain

Information & Communication Technology

Project No.

6406

Sanction and Project Initiation

Sanction No: F. No.: 3-18/2015-T/S-I(Vol. III) 22-03-2017

Sanction Date: 22/03/2017

Project Initiation date: 01/09/2017

Project Duration: 36

Partner Ministry/Agency/Industry

Ministry of Electronics and Information Technology, Government of India

 

Role of partner:

 

Support from partner: Partner Ministry arranged workshop at its New Delhi HQ. where technology needs were identified.

Principal Investigator

PI Image

Goutam SahaPrimary
Indian Institute of Technology Kharagpur

Host Institute

Scope and Objectives

Speech-based biometric systems received a significant attention from research community because of its application in banking, human-computer interfaces, secure access control etc. However, such kind of systems are highly susceptible to spoofing attacks specially for remote access cases. With recent development in voice conversion (VC) and speech synthesis (SS) algorithms, spoofing attacks have become serious potential threats to the current state-of-the-art Automatic Speaker Verification (ASV) systems. To impede such attacks and to enhance the security of the ASV systems, development of efficient anti-spoofing algorithms is essential that can differentiate synthetic or converted speech from natural or human speech. The project aims to develop a synthesized detector using technology gaps in synthesized speech generation process for its use in speech interfaces where voice hacking is an issue. It will also consider real-time use of the counter measure developed in an automatic speaker verification platform when subjected to different kinds of spoofed speech.

Deliverables

Projected outcome is to develop a system and method as application to speech interface devices such as smart phone. The application would differentiate natural and synthetic speech to prevent real time spoofing attack in the speech-based biometric authentication system. The process involves (i)development of data acquisition system that can record speech samples, (ii)design of advanced signal processing techniques to detect spoofing activity which would be put as countermeasure and (iii)integration of automatic speaker verification system with the countermeasure.

 

Scientific Output

We do not want to put the intermediate findings in the public domain. It will be put in due course after securing the IP / peer review. The findings were put before the expert reviewers in the last review meeting.

 

Results and outcome till date

We have developed a desktop based anti-spoofing detection application using open-source programming environment. The application takes the speech as input from the connected microphone, then preprocesses the signal before it is fed to the anti-spoof detection engine of the application.

 

Societal benefit and impact anticipated

Automatic speaker verification (ASV) system is a convenient and cost-effective option for remote authentication of a person using voice based biometric. Advances in this field have improved the ASV performances for which it is getting adopted for banking, e-commerce, personalized device etc. However, in parallel, techniques like voice conversion, speech synthesis have come up which are algorithmic based approaches to generate a target speaker's voice. Such synthesized speech can be used for spoofing attacks and it raises concern on reliability and usefulness of speaker verification systems. The identity hack through synthetic speech can cause huge problem even in our day to day life where we may not be sure about the person at the other side of a telephonic conversation and that would jeopardize our privacy and security. There lies the scope of developing efficient ant-spoofing counter measure and employing it in conjunction with speaker verification systems. Recently held international challenges like ASVspoof, BTAS show the opportunity that lie in this field and the concerted effort of the international research community to come up with an acceptable solution. The research and development initiative undertaken in this project attempts to make the wide spread use of mobile phone more secure against voice based identity hack.

Next steps

We are developing algorithm as was proposed and have also started exploring techniques that were suggested in the last review.

Publications and reports

 

Patents

 

Scholars and Project Staff

One project staff is working in the project who joined on September 01, 2017.

Challenges faced

Lack of manpower due to non-availability of the fifty percent of the allocated 1st and and 2nd year grant. The second project staff joined on Nov 11, 2017 but resigned on Jan 29, 2018 due to Mother's ill health. We could not fill up the vacancy due to fund problem.

Other information

 

Financial Information

  • Total sanction: Rs. 0

  • Amount received: Rs. 0

  • Amount utilised for Equipment: Rs. 0

  • Amount utilised for Manpower: Rs. 0

  • Amount utilised for Consumables: Rs. 0

  • Amount utilised for Contingency: Rs. 0

  • Amount utilised for Travel: Rs. 0

  • Amount utilised for Other Expenses: 0

  • Amount utilised for Overheads: Rs. 0

Equipment and facilities

 

Dell PowerEdge Server, Dell Desktop i5, UPS