This text explores the use of Neo4j, a graph database, for fraud detection. An anonymized dataset of user accounts and transactions from a peer-to-peer platform is analyzed to identify potential patterns of fraudulent activity. The authors introduce the concept of community detection in graphs to partition users into well-connected groups, where the connectivity within these communities is higher than outside them. They apply Louvain community detection to the graph schema and aggregate statistics, revealing suspicious patterns such as flagged users sending money to non-flagged users who share devices, cards, and IP addresses with them. These findings suggest that the receiving user accounts may be fraud risks, highlighting the potential of graph databases in uncovering complex patterns indicative of fraudulent activity.