Skip to content. | Skip to navigation

Personal tools

Navigation

You are here: Home / Sponsors / Collection of voice spoofing database

Collection of voice spoofing database

While vulnerabilities and countermeasures for other biometric modalities have been widely studied, automatic speaker verification systems still remain vulnerable and need further investigation.

One of the most popular applications of voice recognition relates to telephony where the only equipment needed is a regular telephone. In such scenarios the use of other biometrics is not possible without additional and perhaps costly equipment. In addition to others where voice recognition finds utility, this scenario is wholly unsupervised, i.e. there is no human supervision of the acquisition process. Voice recognition is therefore particularly vulnerable to spoofing threats since the only means of detecting the manipulation or tampering of the biometric sample must come directly from the speech signal itself. Such is the concern of the community that several studies have investigated the vulnerability of speaker recognition systems to spoofing.

Most proven threats to voice recognition systems relate to some form of automatic replay or manipulation of natural speech. Replay attacks refer to the situation when an intruder plays back a recording of a true client in order to spoof a voice verification system. Replay attacks, though, arguably present more of a threat to text-dependent systems, i.e. where the text is predictable and where speech signals are regularly repeated and thus easily recorded and replayed. Such attacks are thus of less relevance to text-independent and interrogative or interactive dialogue-based systems where the text is less predictable and rarely repeated. More sophisticated attacks include voice transformation/conversion and speech synthesis.


Apply for affiliateship!

Filed under:
Sponsorship

Help us to push the boundaries of biometrics technology

  • Fund research activities such as data collections, biometrics competitions, etc.
  • Donate to the center to support the infrastructure
  • Make in-kind contribuitons
This is themeComment