A Novel GRU-Attention Framework with Adaptive Authentication for Robust Phishing Attack Detection and Secure Data Transfer

Phishing attacks have become increasingly sophisticated, using advanced techniques to mimic legitimate websites and trick users into revealing sensitive information. Traditional phishing detection models often fail to cope with the evolving nature of these attacks, particularly because they struggle to capture the sequential patterns inherent in phishing transactions. To address this limitation, we propose an enhanced phishing detection framework based on a gated recurrent unit (GRU) with an attention mechanism, designed to achieve more robust and accurate detection. The system integrates adaptive authentication, which applies dynamic security measures to ensure safe data transfer. Unlike previous approaches that focused solely on detection accuracy, our framework introduces a novel integration of GRU + Attention with adaptive authentication, enabling both accurate detection and context-aware protection. Data collection is carried out using the credible OpenPhish source. Preprocessing involves Min-Max normalization, followed by feature extraction with artificial neural networks (ANN), while the Fox optimizer is applied for feature selection to balance exploration and exploitation. Experimental evaluation demonstrates that the proposed model achieves an accuracy of 99.19%, with precision of 99.11% and recall of 99.88%. Compared with existing approaches such as random forest, ensemble learning, and FastText with LSTM, the GRU with attention mechanism consistently outperforms across all metrics. This approach offers a flexible and scalable solution for mitigating phishing threats in dynamic online environments.