TwExodus

Twitter application to mass contact your followers.

Understanding of the problem

With the strong network effects Twitter currently presents as a platform, Twitter will closely prevent any attempt at migrating your user base to another service.

Of course, this protective behaviour is against your interests as Twitter user, along with the difficulty of contacting users without their explicit consent. This last point is really important as there are legal risks associated with this specific aspect of user contact.

Proposed approach

The Mass DM solution seems to be the right path at contacting users, but it must be done in a stealth way. So building stealth techniques to the solution was my main focus at development.

Elaborating on the basic stealth techniques currently implemented there are 3 items that distinguish:

Time randomization: Sleep script between a random amount of seconds per direct message sent
Time window: Do not run after too late in the night, as in simulating an intern doing manual work
Pool of predefined messages: Text file that contains predefined messages, to add noise and avoid detection. I recommend having at least 75-100 unique messages per 10k followers

There's of course a lot of room for improvement for this specific implementation, here are some ideas:

Usage of K8s instances for spawning different IP addresses
Alternating set of authorized API keys
Run at random times of the day with short bursts
Reverse engineering authorized apps and debugging memory space or attack with a man-in-the-middle technique the SSL/TLS traffic to track API calls for exact traffic mimicking (this would require a custom/rooted Android build or a jailbroken iOS device)
Usage of a mobile phone emulator with automated task jobs (i.e. Monkeyrunner) using the official Twitter application
Among several other stealth techniques

While it's possible to use web scrapers (e.g. Selenium) to navigate Twitter, it was my technical choice to avoid the usage of said technology, as in my experience I believe the functionality of web scrapers is far more prone to fail and require constant (almost daily) fine-tuning to work properly. I chose to use Twitter API wrappers because after all, as long as you use API keys from an already authorized application, Twitter should not be able to tell the difference, unless this script is abused mechanically. Pretty much the user has to keep this in mind while using this script.

I didn't explore the Affiliate Link solution as I am very limited in web development skills. With little modification, it can be made compatible with that approach, as the screen_name (username) field is already in scope for string concatenation when you send the private direct message.

A long term solution would be to build proper analytics and community management tools on the top of the simplicity of Twitter. Pretty much like what Klout used to be. Yet, to supply an immediate market need given the existence of the other platforms (Substack, Ghost, Locals and the like) it's best to support the migration of the follower base.

Technical features

Retrieve followers and store these records in the followers table, both in online and offline mode
Mass DM with randomized direct messages out of your pool of predefined messages
Parsing of received email addresses in direct messages

Requirements

Basic knowledge of command line usage
Basic knowledge of virtual environments in Python
You must have consumer_key and consumer_secret from an already authorized application

Instructions

Navigate to the cloned repository directory
Create a virtual environment (e.g. $ python -m venv venv)
Activate the virtual environment (e.g. $ source venv/bin/activate)
Install package requirements (e.g. (venv) $ pip install -r requirements.txt)
Authorize your Twitter account following these instructions: Medium article
Duplicate file tw_config_sample.py into tw_config.py replacing with your custom values
Run the initialize.py script as this will create the empty database
Run the retrieve_followers_online.py script as this will populate the followers table with your Twitter account followers, alternatively you could use the retrieve_followers_offline.py if you have a follower.js file in the input directory
You might want to change priority fields in the followers table with an SQLite 3 editor/client
Customize your direct messages file in input/messages.txt, you can either have affiliate links or you could request email addresses
Run the send_mass_dm.py script whenever you're ready, every day
If you requested email addresses instead of sending affiliate links, you can use the parse_dump_emails.py script to parse email addresses from direct messages and dumping them in a CSV file in the output directory

Implementation

Developed with Python 3.8.3
Works with SQLite 3

Nice to have's for the future

Several sets of API keys
Machine learning message generator, for added noise
Desktop/On premises native application
Markov chain or AI text generator to provide even more variety of messages to send as DMs
Timezone detector of the follower instead of using the timezone of the main account

IvanMontillaM/TwExodus