Audio fingerprinting uses the Web Audio API to process a synthesised signal through the browser's audio stack and measures the tiny floating-point differences in the output caused by the device's DSP implementation, audio driver, and CPU. No audible sound is played; the technique works entirely on in-memory audio buffers.
Every device processes digital audio slightly differently due to variations in the CPU's floating-point unit, the audio driver, and the operating system's audio subsystem. By routing a known signal through the Web Audio API and examining the numeric output, it is possible to derive a value that is consistent for a given device but differs across devices.
Audio fingerprinting is complementary to visual techniques like canvas and WebGL: it operates on a different computational subsystem, so devices that produce similar canvas output may still be distinguishable by their audio characteristics. Combined, these signals increase the overall entropy of the fingerprint.
In doorman-benny
In doorman-benny, the audio signal contributes to the browser-level `fingerprint`, reflecting DSP behaviour that can vary between browser engines even on the same hardware.
Audio signal (docs)Frequently asked questions
Does audio fingerprinting play any sound to the user?
No. The technique synthesises a signal entirely in memory using the Web Audio API and reads back the processed numeric output. No audio is routed to speakers or any output device.
How stable is an audio fingerprint over time?
For a given device, the audio processing output is typically very stable between visits. Hardware changes, OS updates, or audio driver updates can alter the result, but day-to-day stability is generally high.
Can audio fingerprinting be blocked?
Some browsers and privacy tools add noise to the Web Audio API output or return constant values. An absent or uniform audio signal is itself a distinctive pattern that may be used as a detection signal.
