Summary of an article published on The Web Scraping Club on 2023-05-21 (free to read). Read the full article
A device fingerprint is a set of attributes the device and its browser hand over on their own: OS, screen size, user-agent, language, fonts, plugins, IP address, HTTP headers. The more pieces we add, the fewer users share the same configuration, so a site can follow a small niche of users without cookies. That matters now that cookies are refused or wiped, and anti-bot vendors build on the same basics, add Canvas and WebGL results, and look for incompatible data or outliers.
I explain how the data is collected, splitting active from passive techniques, and why marketing and anti-fraud both want it. I also cover the privacy side under European rules. Then we go layer by layer on masking, using deviceinfo.me to see what leaks: the connection layer, with TLS fingerprint, IP and proxy types, and the browser layer, with Playwright, Chrome, and anti-detect browsers like GoLogin.
What’s inside
- Canvas and WebGL work by rendering an image off-screen, which comes out differently on different hardware, while passive techniques only read what the browser and the HTTP layers pass along.
- Why the opinion of the European data protection working party makes the ePrivacy directive apply to fingerprinting, as it does to cookies, and what that means for visitors.
- At the connection layer, IP-derived data like country, state and ISP, and the choice between datacenter, residential and mobile IPs.
- Small machine details that can give away a scraper, such as a server with no microphone or camera, and the computing cost of headful browsers.
In the full article: Screenshots from deviceinfo.me comparing Playwright in headless and headful mode and a GoLogin profile, the full list of collected attributes, and the links to our TLS fingerprinting post and the working party opinion.