Aggregate privacy: k-anonymity + differential privacy
One "team mood" source topic, two live.aggregate exports with
identical reducers. The raw one publishes exact values on every submission.
The protected one declares privacy: { k: 3, epsilon: 1.0, noise:
'laplace', strategy: 'hybrid' }: below 3 distinct contributors in
the current one-minute round it simply does not move - the
last published value is held, never replaced with a null or a marker, because
the drop below k is itself the signal k-anonymity hides. At k it publishes
with calibrated Laplace noise.
Submitting as Snappy Llama - open this page in two more browsers (or incognito windows) to cross k = 3.
How is your day? (each submission feeds both aggregates)
Raw (this round)
exact, every eventNo submissions this round yet.
Protected (this round)
k = 3, Laplace noiseHeld. Nothing has been published: fewer than 3 distinct contributors have fed this window since you subscribed.