Detecting Malicious URLs in Brushing Scams: A Machine Learning Approach with Human-Centered Cybersecurity

Recently, brushing scams have emerged as a notable cybersecurity threat. These scams involve the unsolicited delivery of packages to unsuspecting individuals, often accompanied by deceptive QR codes designed to harvest personal information. Despite the growing prevalence of such attacks, brushing scams have received limited attention in cybersecurity literature and public awareness efforts. This research presents a comprehensive examination of brushing scams reported across the United States, exploring their underlying motives and implications for data privacy. Our primary objective is to identify and mitigate the risks associated with malicious URLs embedded in QR codes linked to brushing scams. To do this, we use various machine and deep learning models like Support Vector Machine (SVM), General Regression Neural Networks (GRNN), Logistic Regression, Gated Recurrent Units (GRU), Bidirectional Long Short-Term Memory (BiLSTM), and BERT. We evaluate these models based on accuracy, precision, recall, and F1 score to determine their effectiveness in detecting harmful URLs. An innovative aspect of our work is the integration of Explainable AI (XAI), which enhances transparency in model predictions and offers insights into decision-making processes. This approach has the potential to inform future research and improve user trust in automated cybersecurity systems. Additionally, our study highlights the critical role of human factors in cybersecurity. We discuss how increasing user awareness and digital hygiene practices can serve as proactive defenses against brushing scams and similar social engineering attacks. By combining technological and human-centric strategies, this research contributes to the advancement of cybersecurity methods aimed at protecting individuals from evolving digital threats.