Audio Fingerprinting

Audio fingerprinting uses the Web Audio API to generate a signal that differs across devices due to variations in audio hardware, drivers, and browser audio processing implementations.

How it works

Neoprint creates an OfflineAudioContext with a triangle wave oscillator connected through a dynamics compressor, renders 5000 audio samples, and sums the absolute values of a specific range of samples.

// Simplified internal logic
const ctx = new OfflineAudioContext(1, 5000, 44100)
const osc = ctx.createOscillator()
const compressor = ctx.createDynamicsCompressor()

osc.type = 'triangle'
osc.frequency.setValueAtTime(10000, ctx.currentTime)
compressor.threshold.setValueAtTime(-50, ctx.currentTime)
compressor.knee.setValueAtTime(40, ctx.currentTime)
compressor.ratio.setValueAtTime(12, ctx.currentTime)

osc.connect(compressor)
compressor.connect(ctx.destination)
osc.start(0)

const buffer = await ctx.startRendering()
const data = buffer.getChannelData(0)

let sum = 0
for (let i = 4500; i < 5000; i++) {
  sum += Math.abs(data[i])
}

The resulting sum value is a floating-point number unique to the audio processing pipeline.

Why it produces unique results

Factor	Effect
Audio hardware	Different DACs and audio chips process signals differently
Sample rate	Hardware-dependent (44100Hz, 48000Hz, etc.)
Compressor implementation	Browsers implement DynamicsCompressorNode with subtle differences
Floating-point precision	CPU architecture affects intermediate math results
OS audio stack	CoreAudio (macOS), WASAPI (Windows), ALSA/PulseAudio (Linux)

Entropy and stability

Property	Value
Entropy	~8 bits
Stability	0.80
Typical duration	10-15ms

Audio fingerprinting was optimized from 44100 samples (1 second, ~120ms) down to 5000 samples (~15ms) without losing accuracy.

Privacy countermeasures

Browser	Method
Brave	Adds noise to audio output (farbling)
Tor Browser	Blocks OfflineAudioContext entirely
Firefox (resistFingerprinting)	Returns uniform audio data

Neoprint's detectNoise() function checks audio output for variance across multiple renders to detect noise injection.

Usage

const fp = await neoprint.get({ collectors: ['audio'] })
const audio = fp.components.audio.value

console.log(audio.sum)         // e.g. 124.043
console.log(audio.sampleRate)  // e.g. 44100
console.log(audio.length)      // 5000