Voice morphing attack blends identities to bypass voice biometrics: study

A new research paper explores a signal-level approach to voice morphing attacks that exposes vulnerabilities in biometric voice recognition systems.

The abstract describes Time-domain Voice Identity Morphing (TD-VIM) as “a novel approach for voice-based biometric morphing” which “enables the blending of voice characteristics from two distinct identities at the signal level.” TD-VIM allows for seamless voice morphing directly in the time domain, allowing “identity blending without any embeddings from the backbone, or reference text.”

“In biometric systems, it is a common practice to associate each sample or template with a specific individual,” the authors say. Advanced Voice Identity Morphing (VIM) enables the generation of a sample that blends the identities of two or more speakers. “The morphed voice sample can be used to match all identities whose voice samples are employed to generate morphing attacks, thus posing a high risk to application scenarios, such as banking and finance, where single identity verification is essential.”

To explore the problem, the research team “created four distinct morphed signals based on morphing factors and evaluated their effectiveness using a comprehensive vulnerability analysis.” Data was benchmarked against the Generalized Morphing Attack Potential (G-MAP) metric, “measuring attack success across two deep-learning-based Speaker Verification Systems (SVS) and one commercial system, Verispeak.”

“Our targeted analysis on Verispeak highlights TD-VIM’s success rate in challenging advanced SVS defenses,” says the conclusion. “The findings underscore TD-VIM’s potential to bypass sophisticated verification measures, emphasizing the importance of enhancing SVS security.”

The research comes out of the Indian Institute of Technology and the Norwegian University of Science and Technology (NTNU).

Article Topics

AI fraud  |  biometrics  |  biometrics research  |  morphing attack  |  voice biometrics  |  voice morphing  |  voice recognition