Studying the spread of militant radicalization through twitter

a data backed approach

Posted by Ankit Shukla on 20th February, 2016

Since the later half of last decade Al Qaeda, ISIS and various other organizations and their affiliates have started embarassing the power of internet mainly exploiting social media due to its massive scale. They use misleading and fearful content to incite fear and the to disseminate their violent propaganda, additionally online platform have also become recruitment channels for these groups. Although these websites are trying their best to ban such users but these people again jump in with new accounts, although the benefit of stopping such account repeatedly is that the new popped up accounts needs an adequate amount of time to build their reach again. This study is based on a list of ISIS operated or affiliated twitter accounts released by xOPERATIONSx, I have applied some data analysis on these users to cruch up some numbers and tried to quantify the overall effect.

The flow of this long post will be

Sources of data:


The list of ~140 ISIS supporting accounts was made public by "anonymous" on 26-27 November, [more lists were made online by the same group but this post is based ony on one list shared by @xoperationsx] (note: although twitter puts forward that the list could be highly inaccurate, the tweets from those accounts definitely show some relation with ISIS). Here is the list of you are so curious
Every other data has been collected by me using fundamental tools which includes twitter api, mysql, php and javascript.
The followers [~35000] of all these accounts are also exhaustively studied as we can safely assume that they are either the immediate target for ISIS recruitment or have already been entrapped.
Total accounts analysed : 140
Out of 140 I could only publically access 110 accounts, that means either the other 30 were taken down by twitter or had strict privacy settings.
To keep the study as recent as possible I studied a maximum of latest 50/30 tweets from each account
This list primarily contained accounts from Indonesia so the study is somewhat biased around this fact.

We considered two types of accounts
Pro-isis accounts : Primary list of so called isis supporter accounts made public by @xoperationsx.
Followers of pro-isis : Accounts which directly follow pro-isis accounts on twitter.

Summary of stats:

Let the graphs take over!

Age of pro-isis accounts on twitter

[serial | made in year | number of accounts]

  1. 2015 49
  2. 2014 26
  3. 2013 13
  4. 2012 10
  5. 2011 6
  6. 2010 5
  7. 2009 1

Age of follower accounts on twitter

[serial | made in year | number of accounts]

  1. 2015 11961
  2. 2014 8702
  3. 2013 5400
  4. 2012 4024
  5. 2011 2659
  6. 2010 1551
  7. 2009 1178
  8. 2008 51
  9. 2007 13
Take Away : Government should work with twitter and other social networks to track these accounts and ban them as often and early as they can, the failure to identify and block them quickly enable them to extend their follower base and spread radicalization swiftly.

Distribution of retweets received by pro-isis accounts

X-axis : Number of retweets
Y-axis : Number of tweets

Distribution of retweets received by follower accounts

X-axis : Number of retweets
Y-axis : Number of tweets

Take Away : The radicals are embarrassing the network effect to full potential, the tweets and content shared by isis is easily reaching the common population (people not directly attached to radicalization groups) on twitter through retweets earned through these ~35000 accounts.

Hashtags used by pro-isis accounts

Hashtags used by pro-isis follower accounts

Take Away : Its apparent from this data that extremist population is using religion based keywords in their communication propaganda, government and intelligence agencies can track these and alike words to track more such accounts and take necessary actions to stop them.

Platforms used by pro-isis to post tweets

X axis sources
Y axis tweets [%]

Platforms used by followers to post tweets

X axis sources
Y axis tweets [%]


Top websites referred by pro-isis accounts

  1. twitter.com 154
  2. fllwrs.com 80
  3. bit.ly 71
  4. fb.me 32
  5. justpaste.it 31
  6. youtu.be 31
  7. panjimas.com 29
  8. syamtodaynews.com 27
  9. youtube.com 23
  10. wp.me 23
  11. azzammedia.net 15
  12. shoutussalam.org 13
  13. zad-muslim.com 12
  14. al-mustaqbal.net 12
  15. pic.twitter.com 12
  16. manjanik.com 12
  17. uapp.ly 10
  18. archive.org 10
  19. ift.tt 9
  20. m.voa-islam.com 8

Top websites referred by Followers' accounts

  1. twitter.com 21190
  2. instagram.com 14159
  3. fb.me 13689
  4. bit.ly 11139
  5. youtu.be 8440
  6. path.com 6053
  7. fllwrs.com 4931
  8. goo.gl 4447
  9. uapp.ly 4370
  10. youtube.com 4161
  11. pic.twitter.com 3254
  12. dlvr.it 2547
  13. quran.ksu.edu.sa 2234
  14. crowdfireapp.com 2139
  15. justpaste.it 2065
  16. du3a.org 1748
  17. chirpstory.com 1542
  18. ow.ly 1271
  19. unfollowspy.com 1183
  20. facebook.com 1023
  21. ln.is 1006
  22. knzmuslim.com 979
  23. wp.me 976
  24. tinyurl.com 795
  25. beasiswaindo.com 758
  26. ask.fm 733
  27. ift.tt 672
  28. shar.es 660
  29. syamtodaynews.com 648
  30. archive.org 630
  31. de.tk 628
  32. swarmapp.com 627
  33. tmblr.co 565
  34. l.ask.fm 554
  35. vine.co 543
  36. kom.ps 537
  37. knz.so 504
  38. manjanik.com 457
  39. m.youtube.com 455
  40. aje.io 394
  41. hizbut-tahrir.or.id 380
  42. tl.gd 379
  43. justunfollow.com 370
  44. rol.co.id 356
  45. bbc.in 353
  46. buff.ly 346
  47. azzammedia.net 343
  48. twitpic.com 331
  49. unfalert.com 325

Languages used in tweets sent from pro-isis

LanguageNumber of tweets
in 1859 ( 64.66 %)
en 571 ( 19.86 %)
und 196 ( 6.82 %)
ar 99 ( 3.44 %)
nl 57 ( 1.98 %)
sl 18 ( 0.63 %)
tl 15 ( 0.52 %)
fi 12 ( 0.42 %)
tr 12 ( 0.42 %)
es 9 ( 0.31 %)
sk 4 ( 0.14 %)
it 3 ( 0.1 %)
fr 3 ( 0.1 %)
hi 3 ( 0.1 %)
bg 2 ( 0.07 %)
ro 2 ( 0.07 %)
is 2 ( 0.07 %)
de 2 ( 0.07 %)
et 2 ( 0.07 %)
ur 1 ( 0.03 %)
bs 1 ( 0.03 %)
zh 1 ( 0.03 %)
lt 1 ( 0.03 %)

Languages used in tweets sent from followers

LanguageNumber of tweets
in 358258 ( 49.31 %)
en 159145 ( 21.9 %)
ar 114303 ( 15.73 %)
tl 7195 ( 0.99 %)
tr 5047 ( 0.69 %)
fr 4575 ( 0.63 %)
es 3873 ( 0.53 %)
sl 2576 ( 0.35 %)
nl 2426 ( 0.33 %)
et 1981 ( 0.27 %)
de 1379 ( 0.19 %)
hi 1325 ( 0.18 %)
ur 1199 ( 0.17 %)
fi 1024 ( 0.14 %)
pt 972 ( 0.13 %)
it 950 ( 0.13 %)
ru 839 ( 0.12 %)
ja 574 ( 0.08 %)
pl 532 ( 0.07 %)
sk 492 ( 0.07 %)
da 465 ( 0.06 %)
ko 413 ( 0.06 %)
no 412 ( 0.06 %)
lt 344 ( 0.05 %)
sv 330 ( 0.05 %)
hu 317 ( 0.04 %)
ro 304 ( 0.04 %)
bn 300 ( 0.04 %)
bs 288 ( 0.04 %)
fa 278 ( 0.04 %)
lv 269 ( 0.04 %)
is 135 ( 0.02 %)
hr 120 ( 0.02 %)
th 119 ( 0.02 %)
vi 117 ( 0.02 %)
am 107 ( 0.01 %)
ckb 83 ( 0.01 %)
ta 76 ( 0.01 %)
iw 60 ( 0.01 %)
ps 59 ( 0.01 %)
zh 56 ( 0.01 %)
uk 42 ( 0.01 %)

Note : The language result is biased as the sample population of [110 accounts] was majorly from Indonesia.

Day wise tweet distribution for pro-isis

Day wise tweet distribution for followers


Distribution follower number of the followers of pro-isis accounts


I get all this interesting data i just started from the 140 accounts available, this shows the beauty when we marry social apis and analytics.


The aim of this post is to show how data can be important to predict and take steps in controlling the activities of radical population on the internet, even when only a very limited set of data is available to start with. Although this post will be a little biased depending on the 140 accounts released by ISIS, the possibilities in real world are endless.

Interested in more? I am available at ankit@in7h.com and looking for some mission driven opportunities in cool companies to work for, This link is my facebook profile & this is my linkedin profile, lets connect the dots!
You many also like to read my previous data analysis on Bollywood movie Drishyam

Want to reward me for this write up and analysis? Just share this in your social networks, that would make me happy for sure :)