DEV Community

John Gakhokidze
John Gakhokidze

Posted on

How bot/troll network works on Linkedin

Building bot/troll network is becoming more and more sophisticated. Below
are some findings on Linkedin, but similar scheme is probably used on other social networks.

Disclaimer:

1. Findings and interpretation reflect our opinion, and may not 100% accurate represent real scenario or actual facts.
2. Tools are provided by 3-rd party, and accuracy is guaranteed to the level that 3-rd service generally provides.
3. We did not use Linkedin API.
4. No PII or any other personal information is stored after analyzes

Steps which are used to build bot/troll network:

  1. There are several accounts created daily from different geographical locations.

  2. Those accounts are using either some dummy company names - when account's country claimed is not in ANZ, NA and EU regions, or associating themselves with big company names where active monitoring of Linkedin membership is not implemented.

  3. Accounts are forming interconnected network - to present themselves as real persons accounts.

Methods used to analyze accounts:

  1. API calls on reaction to owned posts (Not Linkedin API)
  2. Account analyzes on 3rd party posts.
  3. Comments analyzes in AWS Comprehend
Note: "Country" refers to country claimed in account profile.

Findings

  1. Accounts have the highest chance to be bots/trolls mostly active on 3rd party posts, without any or very low personal activities.

  2. Bots/trolls which are active on racial, gender, diversity posts are mostly standalone accounts. Those accounts usually do not provide company name, much personal info. This finding is based on USA, Canada accounts analyzes.

  3. Politically motivated bot/troll accounts usually form network before starting activity.

  4. Politically motivated groups usually do not use all networks at once, but there are periodic activities on 1 network at a time to avoid much attention

  5. Posts by bots/trolls usually return about 40%+ Negative sentiment score in AWS Comprehend.

  6. Accounts from countries where Linkedin is banned have very high chance of being bots/trolls, except Russia, where having Linkedin account is not prosecuted as long as one does not touch politics.

I hope this quick introduction will help to recognize bot/trolls accounts.

Top comments (0)