Entropy (Fingerprinting)

Entropy quantifies uncertainty. A signal that takes many roughly equally likely values across the population carries more bits of entropy than one that is nearly always the same. For example, a rare canvas hash distinguishes a device far more than a common platform string that most users share.

When signals are independent, their entropies add, so combining several moderate signals can yield enough total bits to single out one device among billions. In practice signals are partly correlated, so the effective entropy of a combination is less than the naive sum, which is why careful selection and weighting matter.

Fingerprinting systems estimate the entropy of each signal from observed distributions and prioritize high-entropy, stable signals. There is a tension with privacy and stability: the most identifying signals may also be the ones most likely to change or to be perturbed by anti-fingerprinting tools.