FP-Scanner: when anti-fingerprinting tools make you more unique
Posted 2018.08.15 — Baltimore, USA
Install an extension that lies about your browser and you should be harder to fingerprint. We tested the popular defences. Most of them do the opposite.
The problem with lying
A browser fingerprint is made of dozens of attributes that are supposed to agree with each other: the user agent, the platform, the rendering engine, the fonts, what Canvas and WebGL draw. Tools that spoof some of these attributes leave the others untouched — and the inconsistencies are themselves detectable. FP-Scanner looks for exactly those contradictions.
What we found
- We could detect the use of every defence we studied. Detection alone adds entropy: “runs anti-fingerprinting tool X” is one more distinguishing fact.
- In the worst cases the tools were severely counterproductive — the artefacts they introduce could be exploited to identify the device uniquely.
- Most defences against fingerprinting, however popular, were not effective.
FP-Scanner shaped what came next. It motivated our use of fingerprints for authentication — if inconsistencies betray spoofing, they can also flag an impostor — and laid the groundwork for FP-Tester (automated testing of fingerprinting defences) and FP-Crawlers (bot detection).
Paper. Antoine Vastel, Pierre Laperdrix, Walter Rudametkin, Romain Rouvoy. FP-Scanner: The Privacy Implications of Browser Fingerprint Inconsistencies. 27th USENIX Security Symposium, Baltimore, USA, pp. 135–150. CORE A*, acceptance rate 19%. Paper on HAL · All publications