Detecting Image Spam Based on K-Labels Propagation Model

In order to detect image spam effectively, we propose a method based on a K-labels propagation model (KLPM) in this paper. Specifically, the speeded up robust features (SURF) of each image are extracted firstly. Then to standardize the features of each image, an improved means clustering algorithm is used to cluster these features and get the information of M cluster centers. Finally, after being labeled, all testing images are classified into spam images or ham images via the KLPM, which is based on a K-nearest neighbor (KNN) graph and a label propagating process. Experiments show that the precision of the proposed method can reach to 95%. Therefore, the method based on KLPM achieves a significant improvement in detecting Image Spam.