Technical report on the Content Rating and Fake Detection System

The goal of our content rating and fake detection system is to automatically detect files having paedophile or pornographic content in a P2P system. It evaluates the content of a file using user preferences, the idea being that if a given user is interested in two files, this means that these files are somewhat related. Based on this idea, we build a graph of common interest between files and extract a hierarchical community structure from it, grouping together similar files. Starting from a file list initially considered as paedophile or pornographic, this method characterises all files that are similar to the ones from the initial list, i.e. files with paedophile or pornographic content. In this report, we present the method we implemented and assess the quality of results using various complementary approaches.