A Robust Approach to E-Banking Phishing Detection using Ensemble Methods and LSTM

Recently, the integration of Internet and banking technologies has shifted financial transactions primarily to online platforms. However, this transformation has exposed the financial sector to cyber threats, particularly phishing, resulting in substantial annual losses and widespread data breaches. Despite ongoing efforts to combat phishing, a comprehensive solution remains elusive, creating a vulnerability for cyber attackers to unlawfully access sensitive personal and financial data. In this research, a novel method for successfully recognizing phishing attempts is presented. This proposed methodology includes components such as URLs, HTML, and recurring themes in phishing efforts, which are thoroughly analyzed. These characteristics are consistently similar, resulting in a significant amount of data being collected and sent over the Internet. To preserve the integrity of online financial transactions and shield people from data breaches, a robust phishing attack detection system must be developed. This study applies machine learning techniques such as LSTM, Boosting, Random Forests, and Ensemble models for intelligent phishing detection, attaining a remarkable 97 percent accuracy. To capture complicated data relationships, minimize overfitting, and improve generalization performance, a voting classifier combines predictions and stacking and uses feedback from several models. By adopting this strategy, defenses against the constantly changing cyber threat landscape in the digital financial transaction environment are strengthened.