Exploring GRU-based approaches with attention mechanisms for accurate phishing URL detection

In the dynamic realm of digital advancements, the persistent menace of phishing attacks continues to jeopardize the security landscape for both individuals and organizations. As cyber attacks continue to proliferate, URL-based phishing attacks are growing rapidly. This paper presents an exploratory study aimed at enhancing cybersecurity measures through the detection of phishing URLs. Our approach involves exploring the integration of Gated Recurrent Units (GRU) with various attention mechanisms to bolster accuracy in discerning between legitimate and phishing URLs. Notably, our study reveals that the implementation of the Bahdanau attention mechanism with GRU yields remarkable results, achieving an accuracy of 98.14%. We conducted experiments on a comprehensive dataset comprising 95,913 URLs. Our primary objectives include fortifying cybersecurity defenses against phishing threats, innovating through the integration of diverse attention mechanisms with GRU, and substantiating the efficacy of our model through rigorous evaluation metrics. As the realm of cybersecurity confronts escalating challenges, our research not only offers valuable insights but also charts a promising trajectory for future advancements in cybersecurity strategies.