Replies: 2 comments
-
If Anubis were to require WebGPU, all GNU/Linux users would have to install and use Firefox Nightly just to view websites behind Anubis. Many would need to buy new hardware. Given the current audience of websites using Anubis as of May 2025, I predict this is likely to present an undue burden on human beings viewing websites. Websites using Anubis tend to be made for an audience of developers and users of free software for GNU/Linux systems. These users are more likely to rely on an old computer because it's paid for. They're also more likely to be hardcore software freedom enthusiasts who use an old libreboot-modded laptop that is Respects Your Freedom certified. By "old" I mean things like Technoethical's rebadging of the last generation of ThinkPad laptops before Intel ME became mandatory. Such older computers have tended to lack WebGPU even when used interactively by a human being. In addition, WebGPU still has limited support in web browsers, available in Chromium and not much more. If I try Web Graphics API Tester in Firefox 138.0.1 (64-bit) on Xubuntu 24.04 LTS on a ThinkPad T450 with a Core i5-5300U CPU and HD Graphics 5500 GPU, the browser console shows
|
Beta Was this translation helpful? Give feedback.
-
WebGPU support is out of scope for now. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
This takes advantage of servers used for scraping often not having displays, access to GPUs not even interested ones, nor pointing devices. Touch screens such as those on phones would also count as pointing devices.
If webGPU is detected and fingerprinting detects that it's a bot, then use webGPU for mining crypto before allowing access. The mining process adjusted for difficulty should be completed within a set amount of time according to benchmark data.
Fingerprinting should target modern DevOps, ITOps and cloud practices such as containerization using docker.
Fingerprinting could for example be checking the number of cores or GPU attributes and then doing hashing multithreaded on CPU and GPU checking for latency to make sure it's not virtualized nor containerized, and also doing it while temporarily using device RAM and checking latency or looking at storage latency to make sure the device is real,
fingerprinting will see common fingerprints such as those using docker or server or nondescript CPUs while including GPUs, as bots
https://doi.org/10.14722%2Fndss.2022.24093 describes a way to fingerprint performance of the GPU which maybe could be used to tell apart real machines from docker or virtual machines
Beta Was this translation helpful? Give feedback.
All reactions