What is AudioContext fingerprinting?

AudioContext fingerprinting generates an inaudible waveform, runs it through the Web Audio API's processing graph (oscillator, compressor, and so on), and hashes the numeric output. Small floating-point differences across audio stacks make that hash reasonably stable per device. Nothing is ever played aloud.

As with canvas, naive defenses add per-read noise — which is itself a signal if it changes when a real device's wouldn't, or if a worker and the main thread disagree.

The coherent approach is deterministic per-site farbling (the same site always sees the same subtly-shifted value), derived from the same seed as the rest of the identity so audio agrees with everything else. See browser fingerprinting.