May, 2022: Monthly Report

Posted June 10, 2022 by Aravinth Manivannan ‐ 4 min read

Last Edited June 14, 2022

We are mCaptcha. We build kickass CAPTCHA systems that give (DDoS) attackers a run for their money. And we do all of this without tracking your users. Oh and did I mention our UX is great?

Hello and welcome to the May 2022 edition of the monthly report!

mCaptcha, for a while was showing all the signs of a dead project: no commits on the repositories and no monthly updates. But the project is far from dead!

Python bindings to mCaptcha PoW

pow_py contains bindings to pow_sha256, the proof-of-work library that mCaptcha uses. For the uninitiated, the bindings allow for python programs to automatically solve mCaptcha.

So if you are writing a script to do some chore on your favourite website that is protected by mCaptcha, you can now solve the mCaptcha automatically from within the program.

Here’s an example:

 1import os
 3import mcaptcha_pow_py
 4import requests
 6# get the sitekey that is used in the mCaptcha protected form
 8# the hostname of the mCaptcha instance that the form is using
12GET_CONFIG_ROUTE = f"{MCAPTCHA_HOST}/api/v1/pow/config"
13VERIFY_POW_ROUTE = f"{MCAPTCHA_HOST}/api/v1/pow/verify"
15def solve_captcha():
17    # get challenge configuration
18    key = {"key": sitekey}
19    challenge_config =, json=key)
20    challenge_config = challenge_config.json()
22    # extract configuration data
23    config = mcaptcha_pow_py.PoWConfig(challenge_config["salt"])
24    pow_string = challenge_config["string"]
25    pow_difficulty_factor = challenge_config["difficulty_factor"]
27    # generate work
28    work =, pow_difficulty_factor)
30    # verify PoW
31    proof = {
32        "key": SITEKEY,
33        "nonce": work.nonce,
34        "result": work.result,
35        "string": challenge_config["string"],
36    }
37    resp =, json=proof)
38    resp = resp.json()
40    # extract verification token
41    token = resp["token"]
43    return token
46token = solve_captcha()
47data = {
48    "username": "me",
49    "password": "superlongpassword",
50    "confirm_password": "superlongpassword",
51    "mcaptcha__token": token,
53response ="/mCaptcha-protected-form", data=data)

This could be missed for building DDoS bots(more on that here) but this could also be used to make CAPTCHA solving automated within screen readers and other accessibility devices!

Measuring DDoS protection effectiveness

Proof-of-work has historically been a good method to achieve rate limiting but how much attack can it, specifically mCaptcha’s implementation, withstand when compared to an unprotected endpoint? To find out, we used the recently created Python bindings to the mCaptcha PoW library, the excellent load testing tool, locust and wrote mCaptcha/dos!

VIT AP kindly permitted me, @realaravinth, to use their network security lab for setting up a isolated, contained testing environment to mount a DDoS attack on a test server instance.

The initial topology consisted of one mCaptcha instance, one DDoS demo server, one locust node running in leader configuration and six locust nodes running in follower configuration. I was authorised to use the netsec lab for three days, which unfortunately wasn’t enough to go finish running the experiment. Dr. Sibi Chakkaravarthy Sethuraman has kindly offered to arrange authorisation to use the netsec lab once again in July 2022, during which I hope to finish running the experiment

Special thanks to ackr-8 and alan2000alex for help with setting up infrastructure of the experiment.


mCaptcha underwent a major refactor during the month of May: We re-wrote and cleaned up all database-related stuff for higher flexibility and generally good architecture. This refactor lays the foundation for implementing support for alternate database software programs(we currently support PostgreSQL only).

mCaptcha is now on the Fediverse

We recently joined the Fediverse on a GoToSocial instance run by @realaravinth. We’ll soon be deleting our Twitter account in favour of the Fediverse account.

Fediverse account:

Generic hosting

I, @realaravinth, have been busy with ForgeFlux and Hostea — both of which are software forge related and so when usable, will mostly improve the Free Software ecosystem. Hostea is a project that aims to create a libre software development ecosystem and provide managed hosting for the same. The project is built by a horizontal community, which allows for multiple service providers who adhere to the Hostea policies to operate under the Hostea umbrella — essentially allowing for the creation of smaller, highly localised cooperatives.

Cooperatives are interesting, and we believe that mCaptcha, too, can benefit from such an architecture as it will prevent any one party from single-handedly sabotaging the project. The experience gained from Hostea will be reused in providing managed hosting for mCaptcha.

By the end of this year, mCaptcha will reorganise into a horizontal community and adopt radical transparency to improve trust and reliability of the project

P.S: I, realaravinth, would do it sooner but I’m a little busy right now, so if someone is interested to help out do reach out and so that we could do it sooner!

In context of mCaptcha, radical transparency will include all decisions publicly made, funding and expenses publicly documented, and all collaborations, too, publicly documented. This of course doesn’t imply that private, personally identifiable information(addresses and phone numbers, for instance) will be publicly disclosed. Such information will be redacted and published.

Edit this page on