Opened 9 years ago

Closed 3 years ago

#4147 closed task (wontfix)

What fraction of exit relays exit from a different IP than is in their descriptor?

Reported by: arma Owned by:
Priority: Medium Milestone:
Component: Metrics/Analysis Version:
Severity: Keywords:
Cc: sebastian, karsten Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

If we tell people to use e.g. VisiTor to see how many of their web hits are from Tor, the visitor scripts only look at descriptors and the consensus documents.

Karsten has been working on integrating the tordnsel output with the Exonerator output. How much error is there if you only look at the addresses in the normal directory documents? That is, how much should we care about making sure to keep the tordnsel (or soon, torbel) output integrated?

By fraction I probably mean 'fraction by exit capacity'.

Child Tickets

Attachments (1)

different-exit-address.png (26.3 KB) - added by karsten 9 years ago.

Download all attachments as: .zip

Change History (5)

comment:1 in reply to:  description Changed 9 years ago by karsten

Cc: karsten added

Replying to arma:

If we tell people to use e.g. VisiTor to see how many of their web hits are from Tor, the visitor scripts only look at descriptors and the consensus documents.

Actually, VisiTor only looks at exit lists, not at consensuses or descriptors:

https://gitweb.torproject.org/metrics-utils.git/blob/HEAD:/visitor/HOWTO

Karsten has been working on integrating the tordnsel output with the Exonerator output. How much error is there if you only look at the addresses in the normal directory documents? That is, how much should we care about making sure to keep the tordnsel (or soon, torbel) output integrated?

By fraction I probably mean 'fraction by exit capacity'.

Sounds like a fun analysis. If someone wants to do it and needs help with the data, let me know. Or maybe I'll just do it myself...

Changed 9 years ago by karsten

Attachment: different-exit-address.png added

comment:2 Changed 9 years ago by karsten

"Roughly 15%" is the answer.

Here's what I did:

First, I looked at the consensuses from two weeks in February 2012 to find all relays that got the Exit flag assigned. Second, I summed up the bytes they wrote in the hour after the consensuses' valid-after time. Third, I looked at the exit lists to see if the relay was found to exit via a different address than the relay used as OR address. If there was only a single exit address the day before the consensuses' valid-after time that was different from the OR address, I counted the written bytes as potentially using a different address for exiting.

Here's the result:


The parsing and graphing code is available here.

Does that answer the question?

comment:3 Changed 8 years ago by karsten

Status: newneeds_information

arma, please see my question above.

comment:4 Changed 3 years ago by karsten

Resolution: wontfix
Status: needs_informationclosed

Closing tickets in Metrics/Analysis that have been created 5+ years ago and not seen progress recently, except for the ones that "nickm-cares" about.

Note: See TracTickets for help on using tickets.