<p>In the recent era, social network is a powerful technology to express our views and build the new friendship. But, due to easy process of creating the account in social network, there are number of fake users in the network. Due to the fake users, the fraudulent messages are spread in the network. In this paper, the clustering-based approach is proposed for identifying the fake user in the unattributed static social network. Here, the twitter data set is used to detect the fake users in it. The dataset is represented in graph based on its connections. The nodes in graph are grouped together to form a community using dragon fly clustering. The nodes which are having high Jaccard index are determined as outlier node in the community-based data. Because, this node act as a bridge between the community and it does not belong to any particular community. The optimal feature is detected using Dragon Fly – Neural Network (DF-NN) algorithm by reducing the error rate of the classifier to detect the fake nodes. The proposed method is evaluated using Accuracy and True positive rate with the existing PCA-SVM classification method.</p>