5 ms·

I thought one of the more dangerous threats in Tor was the possibility of someone running compromised exit (or entry) nodes? Why didn't they do that instead of

I thought one of the more dangerous threats in Tor was the possibility of someone running compromised exit (or entry) nodes? Why didn't they do that instead of blocking it?

burgerbrain 15y ago
When you're using Tor, you should be assuming the nodes involved are evil. Chances are most governments already do run their own nodes.
- redthrowaway 15y ago
  Considering the US Navy developed it, and the US Intelligence agencies continue to use it regularly, that's a safe bet.
koenigdavidmj 15y ago
Then you only have a 1/N chance of capturing affected traffic, where N is the (fairly large) number of entry or exit nodes. Also, Tor changes which entry and exit nodes you use on occasion (about every few minutes, if I recall).
- ch0wn 15y ago
  Also, if you communicate through properly encrypted protocols, the exit node won't get anything interesting to see.
  - mike-cardwell 15y ago
    Which is easier said than done. For web traffic, the site has to use https for everything. Secondly, the site has to secure their cookies from being sent over http, either by using HSTS with the includeSubDomains flag, or the "secure" cookie flag. Thirdly, you have to assume that the exit node isn't using a forged certificate from a compromised CA. Fourthly, you have to assume that the entity running the exit node doesn't have a zero day vulnerability for your browser, which they can deliver to you as soon as you make any http request to any site.
- mike-cardwell 15y ago
  The Tor client selects a small list of "Entry Guards" and uses only them as entry nodes. I think this change was made two-three years ago. This was done to prevent the problem whereby anyone using Tor for long enough would eventually connect to all entry nodes, allowing anyone else to set up an entry node to gather a list of most of the persistent Tor users.
kragen 15y ago
If someone runs an entry node, they get to see the IP addresses of 1/N of Tor users, how much data the users send and receive, and when. But they don't get to see what data the users are sending and receiving, or even what exit node they're using. If someone runs an exit node, they get to see (and filter) what data 1/N of Tor users are sending and receiving, and when, but they don't get to see who those users are, or even what entry node they're using. And the data they see can still be encrypted; e.g. connections to Gmail will use TLS/SSL, so even the exit node doesn't get to steal your mail. So that may be why.
- mike-cardwell 15y ago
  FWIW, during the recent DigiNotar SSL shambles, there was evidence that some Tor exit nodes were using forged Google certificates to MITM "secure" connections to Googles services. https://twitter.com/#!/moxie__/status/110863647693221888 https://twitter.com/#!/moxie__/status/110863647693221888
- baltcode 15y ago
  Compiling a small list of Tor users and grabbing un-protected data could still be substantial though.

I thought one of the more dangerous threats in Tor was the possibility of someone running compromised exit (or entry) nodes? Why didn't they do that instead of blocking it?

burgerbrain 15y ago

When you're using Tor, you should be assuming the nodes involved are evil. Chances are most governments already do run their own nodes.

redthrowaway 15y ago
Considering the US Navy developed it, and the US Intelligence agencies continue to use it regularly, that's a safe bet.

koenigdavidmj 15y ago

Then you only have a 1/N chance of capturing affected traffic, where N is the (fairly large) number of entry or exit nodes. Also, Tor changes which entry and exit nodes you use on occasion (about every few minutes, if I recall).

ch0wn 15y ago
Also, if you communicate through properly encrypted protocols, the exit node won't get anything interesting to see.
- mike-cardwell 15y ago
  Which is easier said than done. For web traffic, the site has to use https for everything. Secondly, the site has to secure their cookies from being sent over http, either by using HSTS with the includeSubDomains flag, or the "secure" cookie flag. Thirdly, you have to assume that the exit node isn't using a forged certificate from a compromised CA. Fourthly, you have to assume that the entity running the exit node doesn't have a zero day vulnerability for your browser, which they can deliver to you as soon as you make any http request to any site.
mike-cardwell 15y ago
The Tor client selects a small list of "Entry Guards" and uses only them as entry nodes. I think this change was made two-three years ago. This was done to prevent the problem whereby anyone using Tor for long enough would eventually connect to all entry nodes, allowing anyone else to set up an entry node to gather a list of most of the persistent Tor users.

kragen 15y ago

If someone runs an entry node, they get to see the IP addresses of 1/N of Tor users, how much data the users send and receive, and when. But they don't get to see what data the users are sending and receiving, or even what exit node they're using. If someone runs an exit node, they get to see (and filter) what data 1/N of Tor users are sending and receiving, and when, but they don't get to see who those users are, or even what entry node they're using. And the data they see can still be encrypted; e.g. connections to Gmail will use TLS/SSL, so even the exit node doesn't get to steal your mail. So that may be why.

mike-cardwell 15y ago
FWIW, during the recent DigiNotar SSL shambles, there was evidence that some Tor exit nodes were using forged Google certificates to MITM "secure" connections to Googles services. https://twitter.com/#!/moxie__/status/110863647693221888 https://twitter.com/#!/moxie__/status/110863647693221888
baltcode 15y ago
Compiling a small list of Tor users and grabbing un-protected data could still be substantial though.