[tor-talk] Tor traffic statistics?

Karsten Loesing karsten at torproject.org
Fri Mar 9 10:33:31 UTC 2012

On 3/9/12 10:59 AM, Fabio Pietrosanti (naif) wrote:
> On 3/7/12 10:19 PM, Klaus Layer wrote:
>> "Fabio Pietrosanti (naif)" <lists at infosecurity.ch> wrote on 07.03.2012:
>>> - Which are the first top 20 Tor Exit Node (in terms of amount of Tor
>>> Exit traffic) ?
>>> - Those 20 Tor Exit Nodes, how much bandwidth does respect to the
>>> overall Tor Exit traffic of the Tor Network?
>>> Which is the best way to get data to answer those kind of questions?
>> You can simply go to http://torstatus.all.de, select all exit nodes and sort 
>> the result for Bandwidth. Export to CVS and download it. With this data you 
>> have the top 20 exit nodes and the measured bandwidth for all exits. This data  
>> should allow you to answer the questions. Please let us know the result.
> Ok, i'm seeing it.
> For the bandwidth the data available is "Bandwidth Kb/s".
> I would like to know the historical bandwidth data that's went out of
> exit-traffic "given a period of time" (for example 1 month).

You might like Atlas which has bandwidth histories for different time
periods, including 1 month:


Atlas doesn't allow you to select all nodes with the Exit flag though.
You could identify the largest exit nodes on http://torstatus.all.de and
look them up on Atlas.  (Feel free to open tickets with feature requests
for Atlas on Trac.)

> And, said that "period of time" which was the total amount of bandwidth
> that got trough Tor Exit Node?
> Additionally, when i see "Bandwidth Kb/s" i do not see "Tor Exit
> Traffic" but just "Tor Traffic" that also means:
> - Relay
> - Entry (for Guard nodes)
> Are there Tor-Exit-Traffic-Only statistics?

No.  Relays only report total bandwidth, not bandwidth spent on exit
traffic vs. entry or middle node traffic.

> I would like to provide an answer to the question:
> "How much it would cost to monitor X% of Tor Exit Traffic for an enemy?"
> Where X could be 70-80-90% .
> Now that we have *Big Tor Exit Relay*, and almost all of them are also
> Guard Nodes, the questions is:
> Is it possible that by monitoring just 20 high bandwidth server (for
> example) the enemy can get 70-80-90% of Tor Exit Traffic?
> To be able to provide that answer we would need to have Tor Traffic
> Statistics only for Tor Exit Traffic, provided within a given sample
> period and compared with the Total Tor Traffic, given the same sample
> period.

If you assume that most of the traffic in the Tor network is for exiting
to the Internet and not for directory requests or hidden services, and
if you assume that circuits are on average 3 hops long, roughly 1/3 of
all traffic is exit traffic.  Not a super precise estimate, but should
be sufficient for this case.

Here's a statistic on total relay bandwidth in the network:


> Anyone interested in getting in-depth with such analysis?
> -naif
> _______________________________________________
> tor-talk mailing list
> tor-talk at lists.torproject.org
> https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk

More information about the tor-talk mailing list