FP-Stalker: how long can a browser fingerprint follow you?


A fingerprint that is unique today is not necessarily useful to a tracker tomorrow: browsers update, fonts get installed, screens change. FP-Stalker asks the question that matters — can the same device be recognised again after its fingerprint has changed?

Unicity is not linkability

Four years of Am I Unique data let us separate two things that are often confused. Unicity is whether a fingerprint stands out at a given moment. Linkability is whether successive fingerprints of the same device can be chained over time. We built two linking algorithms — a rule-based one and one based on machine learning. The learned one was more effective; the rule-based one was faster.

What the data says

  • About 26% of the devices in our dataset had fingerprints so distinctive that we could follow them for the entire duration of the study.
  • Paradoxically, these look like privacy-conscious users: browsers configured against cookies and other trackers, which introduces subtleties that set them apart from everyone else.
  • About 20% were hard or impossible to track — unique at any given instant, but too close to many similar devices to be re-identified over time.
  • On average we followed a device for 55 days using nothing but its browser fingerprint.

FP-Stalker became a building block for later work, including DrawnApart, where it links GPU traces to devices.


Paper. Antoine Vastel, Pierre Laperdrix, Walter Rudametkin, Romain Rouvoy. FP-Stalker: Tracking Browser Fingerprint Evolutions. 39th IEEE Symposium on Security and Privacy (S&P 2018), San Francisco, USA, pp. 728–741. CORE A*, acceptance rate 11%. Paper on HAL · All publications