Nowadays, a great part of Internet content is not reachable from search engines. Studying the nature of these contents from a cyber security perspective is of high interest, as they could be part of many malware distribution processes, child pornography or copyrighted material exchange, botnet command and control messages, and so on. Although the research community has put a lot of effort into this challenge, most of the existing works are focused on contents that are hidden in websites. However, there are other relevant services that are used to keep and transmit hidden resources, such as P2P protocols. In the present work, we suggest the concept of Deep Torrent to refer to those torrents available in BitTorrent that cannot be found by means of public websites or search engines. We present an implementation of a complete system to crawl the Deep Torrent and evaluate its existence and size. We describe a basic experiment crawling the Deep Torrent for 39 days, in which an initial estimation of its size is 67 percent of the total number of resources shared in the BitTorrent network.