Abstarct: Phishing attacks are one of the common threats in cyberspace, where attackers deceive users by designing fake websites and steal their sensitive information. Effective detection of phishing websites is a major challenge in the field of cybersecurity. In this research, a new method for detecting phishing websites is presented baxsed on the combination of Fisher Linear Discriminator (FLD) algorithm and Attention-baxsed Gated Recurrent Neural Network (GARU). In the first step, the features effective in detecting phishing websites are extracted and an optimal subspace of features is selected using the FLD algorithm, so that the class separation is increased. Then, the reduced data is fed into a GARU model baxsed on attention laxyers and the classification process is performed. The proposed method is evaluated on a dataset of phishing websites extracted from the Kaggle databaxse and the results show that the proposed model performs better than similar methods with an average accuracy of 99.06%. This research provides an effective step towards improving user security in cyberspace and reducing phishing attacks.