In just a few years, social networking sites have become the most popular landmarks on the Internet. They revolutionized the way we communicate, and socialized the Web. However, while it is now impossible to deny their impact, it can take a variety of forms, not all of them are positive. As a result, the detection of anomalies on social networks is a topic of current research that has attracted researchers since the 2000s. This problem is of crucial importance to prevent abnormal activities. So far, all existing works have been devoted to one-dimensional networks. Our approach attempts to provide a new anomaly detection method based on examining relationships between OSN users using multidimensional networks.