The Sound of Your Hardware
Just as Canvas fingerprinting exploits visual rendering differences, AudioContext fingerprinting exploits audio processing differences. Different sound cards and CPU architectures process complex audio signals slightly differently.
The Mechanics
The script uses the Web Audio API (AudioContext) to generate an audio signal, typically an oscillator or a compressor node. It doesn't actually play the sound through your speakers; instead, it processes the audio mathematically entirely in the background.
Because hardware handles floating-point math and audio compression algorithms differently, the resulting audio buffer varies slightly from device to device. The script hashes this final buffer to create an audio fingerprint.
Mitigation
Like Canvas fingerprinting, the primary defense is noise injection. Privacy-focused browsers add tiny, imperceptible random variations to the audio buffer before returning it to the script, breaking the consistency required for fingerprinting.