Getting started

This package provides a set of utilities and algorithms for online social bot detection based on the paper Social Fingerprinting: Detection of Spambot Groups Through DNA-Inspired Behavioral Modeling

1. Download and setup digitaldna

To install digitaldna, execute:

$ pip install digitaldna

If the installation was successful, and digitaldna is correctly installed, you should be able to execute the following Python code:

from digitaldna.lcs import LongestCommonSubsequence

X = ['banana', 'ananan', 'anana', 'hanoi', 'banas']
estimator = LongestCommonSubsequence()
estimator.fit_predict(X)

2. Sequence your first Digital DNA from Twitter

You can sequence an array of Twitter timelines by simply passing the json filename:

from digitaldna import TwitterDDNASequencer

model = TwitterDDNASequencer(input_file='timelines.json', alphabet='b3_type')
arr = model.fit_transform()

Alternatively, if you have already loaded in memory the json array you can pass it as X parameter:

from digitaldna import TwitterDDNASequencer

X = [[{'user1':'tweet1'},{'user1':'tweet2'}],[{'user2':'tweet1'}]]
model = TwitterDDNASequencer(X)
arr = model.fit_transform()

The arr variable has shape (# of users, 2), the first column will contain the user id and the second the digital DNA

3. Write your first Bot Detector

Once you have your DDNA sequences you can launch the bot detection algorithm with these few lines of code:

from digitaldna import LongestCommonSubsequence

X = ['AAAABBAABABBB', 'AAABAABABBB', 'BAAABAABABBA', 'ABAABBBAB']
estimator = LongestCommonSubsequence()
y = estimator.fit_predict(X)

The y variable will contain a boolean array with the same length of X, where bot accounts are labeled as True and real accounts as False.