PHISHING DETECTION IN SOCIAL MEDIA PLATFORMS USING MACHINE LEARNING

Abstract

Phishing attack is the process where online fraudsters gain unwanted access to internet users’ private information through phishing links. the continuous advancement in technology has lead increase and more sophisticated phishing attacks. phishing attacks continue to pose serious threats to internet users, particularly social media users. different studies have focused on the detection on phishing attempt, but have focused on url-based, attributes, neglecting the possibility of contextual indicators in textual messages.to address these gaps, this study develops a message-based phishing detection model that focused on features such as message length, link presence, and urgency cues. this study used a combination of dataset collected from a nigerian online forum (nairaland), and kaggle phishing dataset, which provides a balanced and diverse sample of both legitimate and phishing messages. the data was preprocessed and used to train three different algorithms which includes random forest (rf), support vector machine (svm), and deep neural network (dnn), and was evaluated using standard evaluation matrices including accuracy, precision, recall, and f1-score. the results from the models evolution shows that the dnn model outperforms all the compared models achieving 99% across all evolution matrices used. this study was further compared with the baseline papers, and the proposed dnn model was able to outperform the baseline model which focused on url-based phishing detection. future work can focused on real-time deployment of the model and integration with mobile messaging systems for proactive defense.