User Protection in Social Networks using Machine Learning Techniques

Open Access
Author:
Zhong, Haoti
Graduate Program:
Electrical Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
February 20, 2019
Committee Members:
  • David Jonathan Miller, Dissertation Advisor
  • Anna Cinzia Squicciarini, Committee Chair
  • George Kesidis, Committee Member
  • Sencun Zhu, Committee Member
  • Dongwon Lee, Outside Member
Keywords:
  • Machine Learning
  • Privacy
  • Cyberbully
  • Social Networks
Abstract:
Huge volumes of data, including images and text, are generated every day as a result of the booming of social networks and their corresponding users. Users share content including their personal information, and spend a lot of time on social networks. As an unexpected side effect, issues such as cyberbullying incidents or private-information leakage have seen a steep rise, and have proven harmful to certain users. Some of the negative effects appear ephemerally while others may have a long term impact on the users. For example, users may get their credit card stolen if they accidentally post a private photo containing such information. Therefore, some protection mechanisms are necessary with respect to these data. However, due to the high cost of manual detection of these harmful incidents, and possible failure for humans to make decisions accurately (some is hidden information or not easily recognizable), it is not feasible to expect human to complete these detection tasks. Machine learning algorithms are suitable to overcome these difficulties we just mentioned and help to provide a better user experience. In this thesis, we explore and design machine learning algorithms to solve three major problems appearing in online social network sites.