negt-dev/funcaptcha-solve-rsa

This project is the decompiled source code for a service that automatically solves FunCaptcha, the visual puzzle challenge made by Arkose Labs and used by many websites to tell humans apart from bots. The author describes it as archived and educational only, noting that it once worked but is no longer operational. The program is written in Go and compiles to a single file with no extra runtime dependencies aside from an outside image recognition service it calls on to read the visual puzzle images. It works by copying the fingerprint of a real web browser, including its TLS and JA3 network signature, so requests look like they are coming from an actual person using Chrome or another common browser. It encrypts data the same way the official JavaScript client does, walks through the same multi step process Arkose uses to issue a challenge, and answers the picture based puzzle waves using its recognition provider before returning a completed token. Inside, the code is split into clear pieces. One part handles the web server and task queue. Another manages the solving steps themselves, including reading the obfuscated site script to recover an encryption key. A separate piece handles proof of work puzzles that some challenges require, either solving them locally or handing them back to whoever is using the service. Other folders manage the browser fingerprint details and the network transport layer, including newer HTTP/3 support. Setting it up means building the Go binary and running it with an API key for the chosen image recognition provider, after which it exposes simple web addresses for starting a solve, checking on it, and handling proof of work steps. The README documents every file and its job in detail, which is unusually thorough for a project of this kind. Because it is marked archived, expect that it may no longer work against the current version of the anti bot service it targets.