Social Network Analysis and Mining Прикладные компьютерные науки

Fake profile recognition using profanity and gender identification on online social networks

Vyawahare, Madhura, Govilkar, Sharvari
Springer Nature · 2022-11-18 · DOI 10.1007/s13278-022-00997-3

Аннотация

Increasing crime rate on online social networking sites is prominently observed in the last few years. Amount of information present and availability of various types of vulnerable people attract criminals on social networking sites. Criminals are misusing the facilities provided by social networks and create false identities which results in identity theft or masquerading, commonly known as clone profile attack and fake profile attack. These false identities are used to carry out numerous types of crimes and hence detection of such attacks is extremely important. Most of the proposed methods are targeting only clone profile attack detection and very few are detecting fake profiles attack. This article differentiates between fake profiles, clone profiles and cross-site cloning and proposes an integrated framework for fake profile and cross-site cloning detection. The proposed solution is integrating fake profile detection with profanity detector module and gender detection to improve the accuracy of fake profile detection. The article considers rich set of impactful features: attribute-based, network-based and user features, and a combination of pre-trained BERT and logistic regression techniques for detection of fake profile attack and cross-site cloning attack with achieved accuracy of 94.65% with precision 88.87% and recall 99.56%. Regression and boosting techniques are analyzed to check which is more impactful in terms of fake profile detection. With this proposed model through experimental analysis, the positive impact of profanity and gender detection on fake profiles detection is identified.

Читать у издателя →