Design and development of user profiling based technique to detect compromised accounts in online social networks
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Security and privacy aspect in social networks has always been a topic of concern, compromisation of legitimate accounts being one of them. Even though with the growing technology, it has become easier to spot and tackle the compromised accounts, but still the rate of hacking of legitimate profiles is comparatively far more than the rate of their detection. Moreover, work on detecting compromised accounts in social networks is still in its infancy and only a few studies focus solely on the detection of such accounts. Furthermore, existing techniques only focus on limited aspects of compromisation, thereby either missing out the actual compromises (false negatives) or falsely predicting an account to be compromised (false positives).In this thesis, different user profiling based techniques have been developed to detect compromised accounts in social networks. The feasibility of authorship verification aka continuous authentication domain have been explored and evaluated for the detection of compromised accounts. To this end, various techniques have been proposed that investigate different features and works on the principle of supervised machine learning for the detection of compromised accounts. Also, unlike existing literature, the problem of compromised account detection has been studied from both binary as well as unary classification perspective. Finally, a multi-dimensional user profiling based technique has been proposed that model features from different dimensions and employ authorship verification procedure to detect compromised accounts in online social networks. Performance analysis and validation of the proposed techniques has been done on the popular microblogging platform called Twitter.
newline