Anti-Spoofing Countermeasure (CM)
Definition
A classifier, also called a CM system, trained to output a score indicating the probability that a given audio segment is genuine or spoofed. CMs are evaluated independently from and in tandem with automatic speaker verification (ASV) systems.
- Function
- Classifies an audio segment as genuine or spoofed speech
- Output
- A score reflecting probability of spoofing
- Evaluation modes
- Standalone, and in tandem with an ASV system
- Applies to
- Voice conversion, synthetic speech, and replay attacks
Common questions
Why is a CM tested separately from the speaker-verification system?+
Standalone testing shows how well the CM alone separates genuine from spoofed audio, while tandem testing shows how the CM changes real-world outcomes once its score is combined with the automatic speaker verification decision, and the two can rank systems differently.
Does a low CM spoof score guarantee the audio is authentic?+
No. A CM is trained against known spoofing methods such as voice conversion and synthesis, so a novel attack technique not represented in its training data can still evade detection even when the reported score looks low.
Related terms
- ASVspoof
- A recurring evaluation campaign and dataset series that benchmarks anti-spoofing countermeasures against corpora of genuine and spoofed utterances. Editions in 2015, 2017,...
- Equal Error Rate (EER)
- An evaluation measure at the operating threshold where false acceptance and false rejection are equal under the specified protocol. It helps compare...
- Neural TTS Cloning
- A text-to-speech system that adapts to a target speaker using a short enrollment recording, generating new utterances in that speaker's voice from...
- Tandem Detection Cost Function (T-DCF)
- The primary evaluation metric in ASVspoof from 2019 onward. It measures the cost of errors when a countermeasure is integrated with an...
- Voice Conversion
- A signal-processing or deep-learning technique that transforms the vocal characteristics of a source speaker's utterance to match a target speaker, while preserving...