The increasing number of Internet users and online shops along with the rise of the number of phishing sites. According to reports from the APWG, the number of phishing reports submitted during the second quarter of 2016 was 466 065. Reports of phishing has increased from 61% at 289 371 received in the first quarter of 2016. Typically, phishing attacks initiated from an e-mail that appears to be sent from a legitimate organization to the victim to update or validate them with the latest information follow the URL link in an e-mail the. In this way the beginning of a phishing attack starts by visiting the link received in an e-mail. It causes significant economic losses, especially for companies. These things underlie research about the classification of phishing sites which will then be classified by two main categories, namely non-phishing sites and phishing. The classification in this study resolved by using NN and KNN. NN widely applied in research because of his ability to model highly nonlinear systems in which the relationship between the variables is unknown (generalization) or very complex, while KNN or K-Nearest Neighbor classification method based on the nearest neighbor distance measurements and has a good performance when given a little training data. Neural networks with backpropagation able to provide the classification of 91.21% larger than using the k-nn with a classification accuracy of 90.33%. Keywords: neural networks, k-nearest neighbor, websites, phishing.