The "dark web" is a term used to describe a network of websites not indexed by traditional search engines and can only be accessed through specific software, such as Tor (The Onion Router). The dark web has gained a reputation for hosting various activities, including illegal marketplaces, forums, and communication channels. In the context of the dark web, researchers have conducted studies to analyze its topology to understand its organization and characteristics of the dark web. This study proposes an approach for community detection in the dark web using a combination of graph theory and machine learning techniques. A graph representation of the dark web is constructed using data collected from Tor dark web. Three popular community detection algorithms, namely Leiden, Louvain, and Label Propagation, are applied to crawled dark web data. The evaluation of these algorithms on a dataset of the dark web demonstrated its effectiveness in identifying meaningful communities. The results obtained from these approaches can provide valuable insights into the organization and structure of the dark web, which can benefit law enforcement agencies and cybersecurity researchers in their efforts to comprehend activities on the dark web.