A SOCKS, HTTP and DNS proxy for distributing traffic across multiple instances of Tor
Go to file
2018-05-11 08:56:11 -07:00
bin Bundles the Tor executable 2018-05-09 22:42:30 -07:00
docs Amends README.md 2018-05-11 08:56:11 -07:00
src Allows signals to be sent to Tor instances using the control protocol from the rpc interface 2018-05-10 22:41:15 -07:00
test allows instances to be removed from pool 2017-03-24 20:11:06 -04:00
.dockerignore removes README from docker build 2017-03-24 20:15:49 -04:00
.gitignore adds control server 2017-03-22 12:45:16 -04:00
.npmignore adds npm ignore 2017-03-24 20:28:58 -04:00
Dockerfile Bundles the Tor executable 2018-05-09 22:42:30 -07:00
LICENSE.txt adds the apache licence 2017-03-25 23:54:41 -04:00
package-lock.json Connects to the Tor instance via the control port when the instance starts 2018-05-10 21:17:00 -07:00
package.json Connects to the Tor instance via the control port when the instance starts 2018-05-10 21:17:00 -07:00
README.md Amends README.md 2018-05-11 08:56:11 -07:00

Tor Router

Tor Router is a simple SOCKS5 forward proxy for distributing traffic across multiple instances of Tor. At startup Tor Router will run an arbitrary number of instances Tor an each request will be sent to a different instance in round-robin fashion. This can be used to increase anonymity, because each request will be sent on a different circut and will most likely use a different exit-node, and also to increase performance since outbound traffic is now split across several instances of Tor.

Tor Router also includes a DNS forward proxy and a HTTP forward proxy as well, which like the SOCKS proxy will distribute traffic across multiple instances of Tor in round-robin fashion. The HTTP forward proxy can be used to access Tor via an HTTP Proxy.

Building and Running

The only installation requirement is node.js. Tor will be downloaded when the node modules are installed.

To install run: npm install To start run: bin/tor-router

To install globally run: npm install -g

Alternatively docker can be used. The build will retrieve the latest version of Tor from the offical Tor Project repository.

To build run: docker build -t znetstar/tor-router . To start run: docker run --rm -it -p 9050:9050 znetstar/tor-router

Usage

The following command line switches and their environment variable equivalents are available for use:

Command line switch Environment Variable Description
-f, --config Path to a JSON configuration file to use
-c, --controlPort CONTROL_PORT Port the control server will bind to (see below)
-j, --instances INSTANCES Number of Tor instances to spawn
-s, --socksPort SOCKS_PORT Port the SOCKS proxy will bind to
-d, --dnsPort DNS_PORT Port the DNS proxy will bind to
-h, --httpPort HTTP_PORT Port the HTTP proxy will bind to
-l, --logLevel LOG_LEVEL Log level (defaults to "info") set to "null" to disable logging. To see a log of all network traffic set logLevel to "verbose"
-p, --parentDataDirectory PARENT_DATA_DIRECTORY Parent directory that will contain the data directories for the instances
-b, --loadBalanceMethod LOAD_BALANCE_METHOD Method that will be used to sort the instances between each request. Currently supports "round_robin" and "weighted".
-t, --torPath TOR_PATH Provide the path for the Tor executable that will be used

For example: tor-router -j 3 -s 9050 would start the proxy with 3 tor instances and listen for SOCKS connections on 9050.

Configuration

Using the --config or -f command line switch you can set the path to a JSON file which can be used to load configuration on startup

The same variable names from the command line switches are used to name the keys in the JSON file.

Example:

 {
 	"controlPort": 9077,
 	"logLevel": "debug"
 }

Using the configuration file you can set a default configuration for all Tor instances

{
	"torConfig": {
		"MaxCircuitDirtiness": "10"
	}
}

You can also specify a configuration for individual instances by setting the "instances" field to an array instead of an integer.

Instances can optionally be assigned name and a weight. If the loadBalanceMethod config variable is set to "weighted" the weight field will determine how frequently the instance is used. If the instance is assigned a name the data directory will be preserved when the process is killed saving time when Tor is restarted.

{
	"loadBalanceMethod": "weighted",
	"instances": [
		{
			"Name": "instance-1"
			"Weight": 10,
			"Config": {
			}
		},
		{
			"Name": "instance-2",
			"Weight": 5,
			"Config": {
			}
		}
	]
}

Control Server

A JSON-RPC 2 TCP Server will listen on port 9077 by default. Using the rpc server the client can add/remove Tor instances and get a new identity (which includes a new ip address) while Tor Router is running.

Example (in node):

	const net = require('net');

	let client = net.createConnection({ port: 9077 }, () => {
		let rpcRequest = {
			"method": "signalAllInstances",
			"params": ["shutdown"], 
			"jsonrpc":"2.0", 
			"id": 1
		};
		client.write(JSON.stringify(rpcRequest));
	});

	client.on('data', (chunk) => {
		let rawResponse = chunk.toString('utf8');
		let rpcResponse = JSON.parse(rawResponse);
		console.log(rpcResponse)
		if (rpcResponse.id === 1) {
			console.log('Three instances have been created!')
		}
	})

A full list of available RPC Methods can be found here

Tor Control Protocol

You can retrieve or set the configuration of instances while they're running via the Tor Control Protocol.

The example below will change the "MaxCircuitDirtiness" value for the first instance in the pool

Example:

let rpcRequest = {
	"method": "setInstanceConfigAt",
	"params": [0, "MaxCircuitDirtiness", "20"], 
	"jsonrpc":"2.0", 
	"id": 1
};
client.write(JSON.stringify(rpcRequest));

You can also send signals directly to instances or to all instances in the pool via the control protocol. A list of all signals can be found here

The example below will set the log level of all instances to "debug".

Example:

let rpcRequest = {
	"method": "signalAllInstances",
	"params": ["DEBUG"], 
	"jsonrpc":"2.0", 
	"id": 1
};
client.write(JSON.stringify(rpcRequest));

Test

Tests are written in mocha, just run npm test