In this paper, stylometric features, gender features and personality features are introduced in the phishing detection. Experimental results show that the detection accuracy reaches 95.05%, which is about 10% higher than that only using the stylometric features. This results indicates that the algorithm has detection effect.