MLSP2018
twitter facebook linkedin share privacy policy contact search

Special Session

DISRUPTIVE RESEARCH IN ANTI-SPOOFING FOR AUTOMATIC SPEAKER VERIFICATION
Research in anti-spoofing for automatic speaker verification has advanced significantly in the last five years. While proposed countermeasures are effective in detecting and deflecting spoofing attacks, current solutions lack a solid grounding in the processes involved in the mounting of spoofing attacks. As a result, and with most current solutions relying on the somewhat blind use of relatively standard features and classifiers, many countermeasures fail when they encounter different forms of attack and are unlikely to generalise well to attacks encountered in the wild. This special session, organised as part of MLSP 2018, seeks to break the mould in anti-spoofing research. We invite scientific contributions that explore fundamentally disruptive approaches to anti-spoofing for automatic speaker verification. While contributions which use existing standard/common databases are welcome, their use is not required. Preference will instead be given to contributions that explore under-researched aspects of spoofing and non-standard, emerging or blue-sky countermeasure technologies, especially those with an emphasis on previously-unexplored signal processing and machine learning approaches which either shed new light on spoofing or expose promising new research directions for future exploration. Both technological and methodological contributions are welcome.

Example topics include but are by no means limited to the following:

  • theoretical bounds of spoofing attack detectability
  • cross-domain feature learning for robust spoofing attack detection
  • generative adversarial networks and threats to biometric technology
  • one-class, semi-supervised, or reinforcement learning approaches to spoofing countermeasures
  • new regularisation and optimisation methods to improve cross-dataset generality
  • generation and detection of inaudible, imperceptible or other novel spoofing attacks
  • novel hardware/sensor and knowledge-based spoofing countermeasures
  • alternatives to GMMs, DNNs, CNNs, RNNs
  • unexpected application areas beyond biometrics
  • Organized by

    Nicholas Evans, EURECOM, France
    Tomi Kinnunen, University of Eastern Finland, Finland
    Sébastien Marcel, IDIAP, Switzerland
    Zheng-Hua Tan, Aalborg University, Denmark

    Submission procedure and schedule is as for regular papers. Please go to paper submission.

    Powered by CONWIZ, © Copyright 2018 | Privacy Policy and Terms of Use | Contact Page Editor