Back to search

University of Sheffield

Characterising and Detecting Social Bots

Abstract

dc:description.abstract

Social bots have been on the rise since the creation of Online Social Networks (OSN). These automated programs bring both positive and negative impacts to our online experience. On the positive end, chatbots can provide instant customer service to us and information bots help broadcast first-hand news. In contradiction, deceptive bots can mimic the real user and used in malicious campaigns such as the romance scam; propaganda bots can spread misinformation and finally swing the election results. A unified approach is proposed to tackle social bot issues in this work. It covered three main stages: First, the very first social honeynet (known as SohoNet) is proposed to monitor and analyse the characteristics and behaviours of social bots. SohoNet is designed to meet three main requirements: accuracy, efficiency and safety. The semi-auto label engine in SohoNet exploits the collective intelligence of multiple honeypots to automate the labelling process partially. Results show that the label engine within SohoNet can auto-label 65% of the profiles captured with high accuracy. On top of that, the overall performance of SohoNet, which is measured based on true positive and capture rate outperformed existing social honeypots. Secondly, an activity-based classifier is proposed to detect partially automated bots. This approach involves classifying profiles activities such as like and follow into natural, automated and mixed. By tracking the profiles on an hourly basis, an interesting dataset is gathered, which reveal the automated-like and follow behaviour. The proposed automated activities classifier has a true positive rate higher than 90%. Last but not least, a ranking model which exploited learning to rank model is proposed to detect social bots in data streams. The solution, known as BotRank ranks users based on their likelihood of being bots. The proposed model achieved high accuracy and recall, which by analysing only the top-k% percentages of users, a significant proportion of bots can be uncovered.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Sheffield
Year dc:date.issued
2020

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Ong, Yew Chuan
Advisors dc:contributor.advisor
  • Ciravegna, Fabio
  • Villa-Uriol, Maria-Cruz

Rights

Language dc:language.dcterms_iso639_2
eng

Identifiers

dc:identifier.*
Identifier
uk.bl.ethos.860637

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Ong, Yew Chuan. Characterising and Detecting Social Bots. doctoral thesis, University of Sheffield, 2020.