DHT Search

This is a Mainline DHT crawler and BitTorrent client which also provides an HTTP interface to query the indexed data.

Distributed Hash Table (DHT) is a distributed system storing key/value pairs, in this case it is specifically Mainline DHT, the type used by BitTorrent clients. The crawler also implements a number of extensions which enable it to get the metadata for the torrent enabling indexing and later searching.

The crawler joins the DHT network and listens to the conversations between nodes, keeping track of interesting packets. The most interesting packets are those where another node announces they have a torrent available.

This BitTorrent client only downloads the torrent metadata. The actual files hosted by the remote nodes are not retrieved.

Features

• Tagging of torrents metadata is fetched. The torrent is tagged using a set of regular expressions matched against the torrent name and the files in the torrent.

• Filtering can be done by tags. By default all torrents tagged 'adult' are not indexed. See the SkipTags option in the configuration file.

• Full Text Search using PostgreSQL's or Sqlite's text search vectors. Torrent names are weighted more than file names.

• Statistics for the crawler process are available when the HTTP server is enabled. Fetch the JSON from the /status endpoint.

• Custom tags can be defined in the configuration file.

• Cross-platform builds for Windows, Macos, Linux, FreeBSD, and OpenBSD

Installation

There is a Makefile for GNU make:

$ make build

Usage

You are going to need to sort out any port forwarding if you are behind NAT so remote nodes can get to yours.

Configuration is done via a TOML formatted file or via flags passed to the daemon.

The following command line flags are available:

  -base-port int
        listen port (and first of multiple ports) (default 6881)
  -debug
        provide debug output
  -dsn string
        Database DSN (default "postgres://dht:dht@localhost/dht?sslmode=disable")
  -http-address string
        HTTP listen address:port (default "localhost:6880")
  -no-http
        no HTTP service
  -num-nodes int
        number of nodes to start (default 1)
  -quiet
        log only errors

and the following "advanced" options:

  -max-bt-workers int
        max number of BT workers (default 256)
  -max-dht-workers int
        max number of DHT workers (default 256)
  -peer-cache-size int
        memory cache of seen peers (default 200)
  -routing-table-size int
        number of remote nodes in routing table (default 1000)
  -tcp-timeout int
        TCP timeout in seconds (default 10)
  -udp-timeout int
        UDP timeout in seconds (default 10)

These options enable you to start a number of DHT nodes thus implementing a small scale Sybil attack. The first DHT node will take the port specified and each subsequent port is for the following nodes.

TODO

• Enable rate limiting.
• Improve our manners on the DHT network (replies etc.).
• Improve the routing table implementation.
• Add results pagination.
• Add tests!