what is fingerprinting?
identifying a client from the combination of properties it exposes, so it can be recognised again even after its ip changes.
a site collects dozens of attributes: header set and ordering, tls handshake details, http/2 settings frames, screen dimensions, timezone, installed fonts, canvas and webgl rendering output, audio processing quirks, cpu core count, language, and how a pointer moves across the page. each one individually is shared by millions of people. combined, they are close to unique and stay stable across sessions and across ip changes. that is the point, and it is why rotating addresses alone does not make you anonymous. two failure modes matter in practice. an inconsistent fingerprint is worse than an ordinary one, so a linux server claiming to be an iphone stands out more than a plain server would. and reusing one fingerprint across many exits links those exits together, so a single flagged account can burn the rest of the pool you were spreading across. exit quality and client plausibility are separate problems; buying proxies only solves the first.