Jump to content


Photo
* * * * * 1 votes

Sniffing traffic on Tor exit nodes


  • Please log in to reply
21 replies to this topic

#1 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 11 September 2007 - 02:38 PM

I would have normally reserved my comments on this for an entry on my blog, but A) I haven't started any topics here yet, B) blogs are already blowing up with content on the subject, and C) there doesn't seem to be any discussion of it here.

A while back, Dan Egerstad of DEranged Security posted a sampling of 100 email accounts, passwords, server IP addresses, and the government and embassy they are associated with. This, understandably, made some people upset. Nearly two weeks later, he posted a small writeup answering the question of "how?". You can read all about it here:

http://www.derangedsecurity.com/ (original source, so you're not getting it from a blog blogging about a slashdot article, or whatever)

To sum things up, he ran 5 Tor exit nodes and simply sniffed the traffic outbound from them. This isn't surprising, or at least it's not supposed to be (I was actually more interested to learn that embassies were using the public Tor network for email traffic like that). It's well-documented that Tor is meant to provide anonymity, though not necessarily a secure channel. Things are encrypted while they bounce around the onion routing, but once it reaches the exit node, the last of these layers of encryption is peeled off, so that the exit node can actually establish and maintain the session to the destination address. Folks on the other end can't tell who the transmission is from (unless the content betrays this information), and that is the purpose of Tor, but they can certainly inspect the content.

However, this does seem to be very surprising to some people. I've seen slashdot stories, ISC posts, and all sorts of discussion of the "attack" of having exit nodes sniffing.

The moral? Understand what part of the security problem is solved by the technology you are using.

#2 gloomer

gloomer

    Hakker addict

  • Members
  • 588 posts

Posted 11 September 2007 - 02:50 PM

Are Tor servers the same as Tor exit nodes? If not, how did he set himself up with 5 Tor exit nodes?

I'm assuming that traffic is heavily encrypted on regular Tor servers, but has to be unencrypted for the exit nodes?

#3 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 11 September 2007 - 06:14 PM

Are Tor servers the same as Tor exit nodes? If not, how did he set himself up with 5 Tor exit nodes?

I'm assuming that traffic is heavily encrypted on regular Tor servers, but has to be unencrypted for the exit nodes?


To anonymize the connection, it is forwarded through a number of Tor onion routers, each "hop" of which is encrypted. This is done in such a way that the intermediate hops through the Tor network can't see the content. This is for the sake of protecting the intermediate hops from being responsible for the content that might be flowing around, and, to some extent, limit where unencrypted traffic can be seen. This encryption is done in layers for each hop (thus the "onion" in onion routing). The details of the protocol are explained much better on the Wikipedia entry and the official Tor site.

Exit nodes are nodes that have volunteered to perform the final hop, which is to decrypt the last layer of encryption over the traffic, and make the connection to the client's destination. This must be done without the encryption of the Tor network, because the client's destination isn't talking the Tor protocol anymore (you can have hidden services that terminate within the Tor network, though). So, anyone can be an exit node, all it takes is some configuration changes, and you'll be added to the list of available exit nodes.

This may seem "wrong", just letting anyone perform the role of exit node. It does improve the anonymity of the network to have a large number of exit nodes (and would greatly decrease the network's effectiveness if it were limited to "trusted" nodes, even if you could define such a thing). It is up to the user to realize the threat posed by an exit node, and prevent it from being an issue by using end-to-end encryption (such as SSL or SSH) with the destination, when possible. With careful reading of the Tor FAQ, documentation, and a healthy dose of paranoia, one can have a pretty good level of anonymity when using Tor, and keep themselves secure in process.

#4 livinded

livinded

    Dangerous free thinker

  • Agents of the Revolution
  • 1,942 posts
  • Location:~/

Posted 11 September 2007 - 07:11 PM

This is one of the reasons I stopped using tor, in a lot of ways it can be a lot more insecure than just using the random place you are at. There are people actually watching the tor network looking for interesting traffic and you never know who has access to the tor exit nodes. Tor in and of itself is a good idea, the only downside is that you need to trust the servers and the people running them.

#5 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 11 September 2007 - 07:27 PM

It's fine for browsing around sites that you wouldn't want to show up in the logs of, and for services that you are certain are going over SSL. You're basically dealing with the same threat model as using an unsecured wireless access point: You have to assume that someone's listening. You should secure the communications you can accordingly, and keep the threat.in mind for the communications you can't.

As many hops as you take over the public internet for non-SSL passworded sites (like this one!), it'd be good practice to assume someone's listening there too. Binrev's about 20 hops from where I'm sitting right now. Each represents a point at which a curious admin or an attacker with a compromised system could be watching my traffic. What can I do about that? Not much. I can mitigate it by using a unique password and keeping an eye on my own post history though ;)

#6 wirefire

wirefire

    HACK THE PLANET!

  • Members
  • 62 posts
  • Location:England

Posted 12 September 2007 - 04:43 AM

Does anyone know if there are any plans to support binrev over https?

#7 Zapperlink

Zapperlink

    "I Hack, therefore, I am"

  • Agents of the Revolution
  • 951 posts
  • Country:
  • Gender:Not Telling

Posted 12 September 2007 - 03:34 PM

Does anyone know if there are any plans to support binrev over https?



Not that I am arguing the benefits of SSL, but why exactly do you feel the urge to need a https session with binrev?

#8 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 12 September 2007 - 03:47 PM

Not that I am arguing the benefits of SSL, but why exactly do you feel the urge to need a https session with binrev?


The primary arguments for this would be to secure the authentication of users and to provide some privacy to things like messages sent between users and exclusive forums. Presumably, nobody here is dumb enough to be using the same password for this site as anything else (hint hint hint hint), so the accounts themselves would be of limited use to an attacker, other than whatever damage could be caused by impersonation. As for private communications between users, it wouldn't be wise to send any messages to other users that you wouldn't want logged and seen by anyone anyways, even if there was SSL for your connection.

I personally, don't see a dire need for it. Admins, and members of the "Agents of the Revolution", which I believe have their own private forums on here, would have a stronger case for wanting an SSL connection, as those accounts would be more desirable to gain access to or sniff content.

#9 StankDawg

StankDawg

    same old Dawg, no new tricks

  • Moderating Team
  • 8,075 posts
  • Country:
  • Gender:Male

Posted 12 September 2007 - 05:05 PM

Yes, I actually did recently purchase a legitimate SSL cert. We have had a self-signed one that we use internally for admin purposes, but not one that we can roll out to the forums yet.

With upwards of 16000 unique visits per day, and an average of 10-12k per day, I worry about the stress that it will cause on our server for that encryption overhead. We are researching this now to determine the feasibility. Unfortunately, the forums do not officially support SSL only for login according to invision, which is teh most important need that I see.

#10 livinded

livinded

    Dangerous free thinker

  • Agents of the Revolution
  • 1,942 posts
  • Location:~/

Posted 12 September 2007 - 09:01 PM

Why not just allow people to choose whether or not they want to use SSL. Personally when I am at home or on a network I trust I'd rather do everything in clear text because I really don't need the extra protection or want to deal with the overhead. The only time I would want to use the SSL would be on an insecure network I don't trust.

#11 m0untainrebel

m0untainrebel

    HACK THE PLANET!

  • Members
  • 60 posts

Posted 13 September 2007 - 02:28 AM

i think everyone piece of online communications should be encrypted, if possible. i mean, why not? the overhead isn't even a factor- it doesn't effect performance at all. there have been times when i've been at a coffee shop using an open network and wanted to login to binrev, and did, unencrypted. for all i know my password has already been sniffed.

i think there's a really big problem with certificate "authorities" getting to hand out ssl certs for money, and that we should actively be seaking an alternative, free, decentralized way of making sure everyone's websites can be secure (like a pgp-style web of trust).

the "if you're not doing anything wrong then why worry" argument is stupid. not to be overly paranoid, but there are all sorts of malicious people eavesdropping on digital communications all the time -- especially on political activists -- and most of the malicious eavesdroppers are cops. if, for example, everyone used public key encryption for their email, no matter how mundane the message, you're helping to create a larger haystack, making all the needles out there that much more secure.

as of now, there's no way to have encrypted communication between your computer and a web server without ssl installed on the server. a lot of people who are worried that the network they're using isn't secure and that there might be sniffers on it use tor for the encryption, but that just pushes your traffic to some exit node to get monitored (which, in some cases, might be better than getting monitored locally...). but anyway, the solution to this problem is to just encrypt everything. that way tor will actually be useful for hiding what websites your looking (anonymity, what it's made for), and ssl will be used for secure encryption (what it's made for). right now, while there isn't a web of trust system built into popular browsers, cacert.org seems to be the best way to get certificate signatures for free- though it's not an authority in big browsers yet either, they seem like they're the most likely candidate.

#12 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 13 September 2007 - 09:20 AM

I think you'd be surprised how much of a performance hit SSL can be to a server.

There is no way you can know "that most of the malicious eavesdroppers are cops". Monitoring by state-sponsored entities is a serious concern, however a statement like that can't really be measured or quantified. "Cops" would have a hard time outnumbering all the civilians out there running Kismet and Wireshark.on networks they can see or direct traffic on.

You're right on with most of that though. It's the usual arguments that cryptography advocates make. I don't know that it'll ever become ubiquitous, as you want, but I do think there's a good message in this that individuals can take away: When you are using a network, be aware of what protections are in place and take responsibility for your own security. There's a lot that the end-user can do to greatly improve their privacy and security, and it's good practice to do so. That unencrypted wireless at the coffee shop might be horribly insecure, but if you're prepared, you can at least make your own communications safer on it.

Since you mentioned certificate authorities, and some folks here might be interested in reading about it, I'll post a link to a blog post I made yesterday about people who give a lot more credit to something being signed than they should:

Thoughts on signed malware

#13 xyzzy

xyzzy

    SCRiPT KiDDie

  • Members
  • 25 posts
  • Location:USA

Posted 13 September 2007 - 03:53 PM

Monitoring by state-sponsored entities is a serious concern


Not really. State-sponsored entities are likely monitoring for benevolent purposes, while rogue entities are likely monitoring for nefarious purposes.

#14 StankDawg

StankDawg

    same old Dawg, no new tricks

  • Moderating Team
  • 8,075 posts
  • Country:
  • Gender:Male

Posted 13 September 2007 - 05:49 PM

I think you'd be surprised how much of a performance hit SSL can be to a server.


Exactly. With the number of users that we have, if they all started using SSL it would cause quite a strain on the server.

#15 m0untainrebel

m0untainrebel

    HACK THE PLANET!

  • Members
  • 60 posts

Posted 13 September 2007 - 06:10 PM

Monitoring by state-sponsored entities is a serious concern


Not really. State-sponsored entities are likely monitoring for benevolent purposes, while rogue entities are likely monitoring for nefarious purposes.


it's true that most malicious eavesdroppers might not be cops (you're right McGrewSecurity, there's no real way of measuring it), i wholeheartedly disagree that state-sponsored entities are benevolent. take the "dark web" for instance: http://yro.slashdot....7/09/12/1728238

there are two reasons why cops (local and federal) spy on people: 1) the victim of the spying is planning on committing a crime or has committed a crime, and 2) the victim of the spying is some sort of political dissident. i think that #2 is always malicious (and it happens more often than people realize), and #1 is malicious quite a bit of the time (monitoring university traffic to track down the p2p users, for example, is malicious behavior).

#16 McGrewSecurity

McGrewSecurity

    SUPR3M3 31337 Mack Daddy P1MP

  • Agents of the Revolution
  • 338 posts
  • Location:Starkville, MS

Posted 13 September 2007 - 06:23 PM

(monitoring university traffic to track down the p2p users, for example, is malicious behavior).


I was following you up to here. Network admins have every right to monitor traffic to track down problems that affect the fair distribution of bandwidth among students, enforce policies on network usage, and prevent illegal activity. If left unchecked, peer-to-peer traffic can saturate the network's bandwidth, and cause the school to receive cease-and-desist letters from copyright holders. You'll find that most schools have policies against abusing network resources in this way, and rightfully so. It's their network, they make the rules.

#17 m0untainrebel

m0untainrebel

    HACK THE PLANET!

  • Members
  • 60 posts

Posted 13 September 2007 - 07:08 PM

my point was filesharing is perfectly ethical, but still illegal. those who try to stop filesharing (the mpaa, riaa, governments, etc.) are malicious, in my opinion. but that's a whole different conversation. it's true, networks can get in trouble for breaking laws, and being a bandwidth hog isn't cool, so bad example.

i think the important thing to remember is that laws and ethics have very little to do with each other. there's nothing ethically wrong with hitchhiking, but it's illegal. there is something ethically wrong with clearcutting oldgrowth forests, but it's legal. every time cops enforce a law that's morally wrong, they're being malicious (even if they are just doing their jobs).

#18 kitche

kitche

    Hakker addict

  • Members
  • 549 posts

Posted 15 September 2007 - 05:43 PM

You really shouldn't try to argure with McGrewSecurity He knows much more about university networks among other university computer related things


#19 WhatChout

WhatChout

    Dangerous free thinker

  • Members
  • 814 posts

Posted 16 September 2007 - 05:12 AM

State-sponsored entities are likely monitoring for benevolent purposes

Just out of curiosity, what makes you think so?

#20 i0null

i0null

    the 0ne

  • Members
  • 1 posts

Posted 16 September 2007 - 09:51 AM

i think everyone piece of online communications should be encrypted


I agree. I can understand the argument that the overhead of the communication, can be significant. However, being more secure often makes it's less convenient. If you only encrypt certain traffic, (like credentials) this also singles out the important traffic, to a potential sniffer.




BinRev is hosted by the great people at Lunarpages!