It's clear, with special configs of this kind, anonymity will suffer. But if one is aware of that and willing to accept that...

How can one make Tor faster?

How can one use only fast relays? Will that actually help?

  • 5
    Why using Tor then? One could go use a random VPN or an open proxy, if not concerned about anonimity/privacy. – mrphs Sep 30 '13 at 15:37
  • It's in the middle. Safer than single hop (logging) VPN, free, not as safe as stock Tor. – adrelanos Oct 2 '13 at 16:16

Tor normally tends to prefer fast relays anyways, so I wouldn't worry about that. The following settings from the Tor Manual can speed things up a bit when tweaked:

ExcludeSingleHopRelays 0|1

This option controls whether circuits built by Tor will include relays with the AllowSingleHopExits flag set to true. If ExcludeSingleHopRelays is set to 0, these relays will be included. Note that these relays might be at higher risk of being seized or observed, so they are not normally included. Also note that relatively few clients turn off this option, so using these relays might make your client stand out. (Default: 1)

This will esentially allow you to use Tor like a normal 1-hop proxy, however, you will essentially lose all anonymity. To actually construct single hop circuits, you'll need to modify the source and rebuild. Check src/or/or.h and change the line:




This is a terrible idea. Don't do it (though it sounds like this isn't what you were looking for anyways based on the comments).

You can also tweak the following settings and may be able to get some speed boost:

NewCircuitPeriod NUM

Every NUM seconds consider whether to build a new circuit. (Default: 30 seconds)

EnforceDistinctSubnets 0|1

If 1, Tor will not put two servers whose IP addresses are "too close" on the same circuit. Currently, two addresses are "too close" if they lie in the same /16 range. (Default: 1)

AllowDotExit 0|1

If enabled, we convert "www.google.com.foo.exit" addresses on the SocksPort/TransPort/NATDPort into "www.google.com" addresses that exit from the node "foo". Disabled by default since attacking websites and exit relays can use it to manipulate your path selection. (Default: 0)

AllowDotExits doesn't do anything by itself (except make you more vulnerable to lots of attacks), but it will allow you to select a fast exit and use it for sites that would suck down bandwidth and be rather slow.

OptimisticData 0|1|auto

When this option is set, and Tor is using an exit node that supports the feature, it will try optimistically to send data to the exit node without waiting for the exit node to report whether the connection succeeded. This can save a round-trip time for protocols like HTTP where the client talks first. If OptimisticData is set to auto, Tor will look at the UseOptimisticData parameter in the networkstatus. (Default: auto)

And many more. Note that messing with these settings is not recommended.

  • 1
    I don't think ExcludeSingleHopRelays does what you suggest it does. Also, users probably shouldn't mess with NewCircuitPeriod, nor am I quite sure why you think it would help. EnforceDistinctSubnets seems dangerous to suggest and probably doesn't help most of the time. AllowDotExit opens clients to all sorts of attacks, and it won't help performance. OptimisticData already defaults to 1 (via auto). – weasel - Peter Palfrader Oct 2 '13 at 19:22
  • @PeterPalfrader Yes, most of these are unsafe, but that's what he was asking for. You're right, just setting ExcludeSingleHopRelays to false doesn't allow single hop's; I'll fix that. Thanks. – Sam Whited Oct 2 '13 at 20:12
  • There's a good discussion on why most of these answers are incorrect here: lists.torproject.org/pipermail/tor-talk/2013-August/029729.html – mttpgn Oct 3 '13 at 20:31
  • That discussion doesn't seem to actually explain its reasoning; someone just says "it doesn't help" a few times. If nothing else several of these settings may reduce network latency (using multiple nodes on the same subnet, for instance). I'll go throw a few of these in a config if I have time later and see if I can get any statistically significant measurable decrease in latency; I may, or I may not. – Sam Whited Oct 3 '13 at 20:33
  • AllowDotExit has been removed: trac.torproject.org/projects/tor/ticket/23426 – nopara73 Aug 5 '18 at 8:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.