Abstract: A long short-term memory (LSTM)-based phishing email detection method was proposed.This method was arranged mainly with two parts:data expansion part and model training part.In the data extension part, KNN and K-means algorithms were used to extend the training data set to make the number of data sets capable support deep learning algorithms. In the model training part,the data were preprocessed and transformed into a word vector matrix.And then the word vector matrix was trained to form LSTM neural network model.Finally,the mail can be divided into normal mail and phishing mail according to the trained LSTM model.Experiments were carried out to evaluate the proposed algorithm.The experimental results show that the proposed algorithm can achieve the accuracy up to 95%.