Investigating Cyber Security Scammer Detection Model for Online Social Network Frauds Using Machine Learning

Abstract

Due to the huge popularity of online social network (OSN), the risk of social network scams and frauds has also escalated. Scammers exploit these platforms by creating scam profiles to deceive unsuspecting users and engage them in fraudulent activities. Developing effective scammer detection algorithms that possess the capability to identify and put an end to various forms of online social network scams such as dating scams, compromised accounts, and scammer profiles detection holds significant importance. This study aims to investigate and develop a robust scammer detection model that leverages machine learning techniques to identify and prevent online social network frauds (OSNF). newlineThe study focuses on two main approaches for scammer detection: Multi-Modal scammer detection and EnsembleScamDetect Multi-Classifier approaches. The Multi-Modal model employs a range of machine learning algorithms like Naive Bayes, Random Forest, Decision Tree, and Support Vector Machine. By examining user profile attributes, these algorithms are used to spot scammers. EnsembleScamDetect strategy integrates multiple classifiers to improve the precision and dependability of scammer identification with various profile features. The research objectives include studying and analyzing OSNF, detecting scammers at an early stage based on user profile attributes, designing and developing a scammer detection model, and evaluating the strength and potential of the model with specific OSNF. The efficiency of the models is evaluated using metrics such as accuracy, precision, recall, and F1 score. The findings of the study indicate that both approaches show promising results in identifying and differentiating between scammer and genuine user profiles. The Multi-Modal model, which utilizes a traditional machine learning algorithm-based technique, attains an accuracy of 94.55%. The EnsembleScamDetect Multi-Classifier approach demonstrates even higher accuracy, with an accuracy rate of 98.75% and an F1 score of 97.95%. These results highlight the ef