Audio Fingerprinting Explained

A common misconception first

Audio fingerprinting does not listen to anything. It never touches your microphone and never plays a sound. It measures how your browser's audio processing engine behaves, using the Web Audio API — the same API music and video applications use to generate and filter sound in the browser.

How the technique works

A script builds a short audio graph: typically an oscillator generating a fixed tone, passed through a dynamics compressor, rendered into a buffer by an OfflineAudioContext. "Offline" means the signal is computed as fast as possible into memory rather than played through speakers. The script then reads the rendered samples — thousands of floating-point numbers — and summarizes them, for example by summing a slice of them.

The input is identical everywhere, so in a perfect world the output would be too. In practice, audio stacks differ: floating-point rounding, compressor implementations, resampling behavior, and hardware paths introduce tiny, consistent variations. The summed output on one system might differ from another's in the sixth decimal place — invisible to any listener, perfectly measurable by code.

Why it works silently

The Web Audio API requires no permission because generating and processing audio math is considered harmless — and for legitimate uses, it is. That is exactly what makes the technique attractive to fingerprinting scripts: it runs on any page without a prompt, completes in well under a second, and produces a stable value.

What this site's test does

The analyzer renders one second of a 10 kHz triangle wave through a dynamics compressor in an offline context, sums a fixed slice of the output samples, and hashes the result locally. The card shows the numeric signature and the sample rate. As always, the value is displayed to you and held only in page memory for comparison — never uploaded, never stored.

Distinguishing power in context

Audio output tends to cluster: machines sharing an operating system version and audio hardware family often produce identical numbers. That is why the analyzer rates this test Moderate rather than Higher — it separates groups of systems well, but within a group it may not separate individuals. Its real value to trackers is as one ingredient among many; combined with canvas and font results, it sharpens the overall picture.

Defenses

Browsers with fingerprinting protections can add small random noise to Web Audio output so the measured numbers change between sessions or sites — Firefox's strict protections and Tor Browser take this approach, and Brave randomizes audio output per site. You can confirm any of these with the analyzer's before-and-after comparison: if the audio hash changes across runs with protection enabled, the noise is being applied.