Securing Personal Data on Smartphones: Addressing APT Challenges and Phishing Threats

Abstract

The widespread integration of smartphones into daily life has revolutionized communication, work, and information access, but it has also made them prime targets for cybercriminals. One significant threat comes from Advanced Persistent Threats (APTs), which involve sophisticated, prolonged cyber intrusions. A common attack vector for APTs is phishing, where victims are deceived into clicking on malicious URLs delivered through SMS, email, WhatsApp, and phone calls. These URLs lead users to cloned websites that look like authentic platforms, deceiving them into disclosing sensitive information including login passwords and personal details. This study examines smartphone security vulnerabilities, with a focus on URL phishing attacks. Our research is organized into two major areas. First, we create a new dataset and use a rigorous feature extraction approach. Second, we present a robust mitigation strategy based on deep learning techniques. Our methodology uses three deep learning models: Convolutional Neural Network (CNN), Deep Neural Network (DNN), and Long Short-Term Memory (LSTM)—each assessed for its effectiveness. The findings underscore the importance of a well-curated dataset and careful feature selection in achieving high performance. The DNN model demonstrated the highest accuracy, the CNN excelled in true positive rate, and the LSTM provided balanced performance. Compared to traditional methods, these deep learning models significantly enhance the detection of phishing attacks, highlighting the crucial role of high-quality datasets in improving model accuracy and robustness.