HHT–SVM: An online method for detecting profile injection attacks in collaborative recommender systems

Collaborative recommender systems are known to be particularly vulnerable to profile injection attacks, in which malicious users insert fake profiles into the rating database in order to bias the systems’ output. To reduce this risk, a number of methods have been proposed to detect such attacks. However, almost all of them operate in batch mode, i.e., they require examining and processing the entire rating database to detect the attacks. With this problem in mind, we propose an online method (called HHT–SVM) to detect profile injection attacks by combining Hilbert–Huang transform (HHT) and support vector machine (SVM), which can operate incrementally. The underpinning idea of HHT–SVM is the feature extraction method based on an individual user profile. In this paper, we first construct rating series for each user profile based on the novelty and popularity of items. Then, by introducing HHT we use the empirical mode decomposition (EMD) approach to decompose each rating series and extract Hilbert spectrum based features to characterize the profile injection attacks. Finally, we exploit SVM to detect profile injection attacks based on the proposed features. We conduct experiments on the MovieLens 1M dataset and compare the performance of HHT–SVM with PCA-VarSelect and Batch-SVM to demonstrate the effectiveness of the proposed approach.