ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Social media user identity linkage by graphic content comparison

Journal: Scientific and Technical Journal of Information Technologies, Mechanics and Optics (Vol.21, No. 6)

Publication Date:

Authors : ;

Page : 942-950

Keywords : social media; user identity linkage; image processing; machine learning; social engineering attacks;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

The article proposes a new approach to comparing accounts of the social media “VKontakte” and “Instagram” to determine those accounts which belong to the same user. The approach is based on the comparison of graphic content; the novelty of the approach consists in combining several methods for matching graphic content, also for the first time a method is proposed for matching accounts of the mentioned social media. The proposed method combines three methods of matching graphic content: by extracting the faces of the account users from the photos in the account and matching them, by matching all faces in both accounts, and by pairwise comparison of images to determine the same images in both accounts using the perceptual pHash method. The described method was tested on a dataset of more than 8,000 pairs of accounts. According to the results of the experiment, the value of the F1-score metric reached 0.87. The practical significance lies in automating the comparison of user accounts in various social networks by implementing of the developed algorithm in the prototype of the software package. A further direction for research lies in expanding the set of data and attributes of profiles considered for comparison. The results can be introduced into a software package for the analysis of the security of a user of information systems against social engineering attacks. It seems promising to combine the obtained findings with account matching methods based on the structural similarity of social graphs.

Last modified: 2022-01-10 20:06:46