Tor: A JAP Replacement

kid_wonder writes "Wired is running an article describing an answer to this previous /. story. Packets are sent through a network of randomly selected servers each of which knows only its predecessor and successor. Packets are unwrapped by a symmetric encryption key at each server that peels off one layer and reveals instructions for the next downstream node. As a 'connection-based low-latency anonymous communication system,' Tor seems to be the answer to JAP to allow anonymous networking activities of all kinds."

Comments

Before you know it...

[cytoman]

... the RIAA and the MPAA will be all over this, denouncing it and crying foul!

sigh...

Freenet?

[pope nihil]

Isn't this onion routing thing exactly what freenet uses?

Re:Freenet?

[Sgs-Cruz]

Which is your right, obviously. But don't be accusing anyone who uses it of trading child porn. I was using Freenet a while back just for the novelty of trying it out -- I found it (much like the Gnutella network) unusable for downloading music/movies so I stopped using it.

But my point is just because it can be used for bad purposes does not mean it necessarily will.

Re:Freenet?

[HeghmoH]

What a crazy attitude to have. There are other reasons you'd need that much anonymity.

First, your dismissal of people who live in China is incredibly inappropriate. Over a billion people live there, and you just dismissed them out of hand. And then there's the exile situation; what about somebody who's now living in the US who still can't speak out freely because of repercussions on friends/family back home? Do they simply not count?

There are plenty of other reasons, though, all the way from "VP in Fortune 500 company wants to expose toxic waste problems without risking being found out as the source" to "I'm such an incredibly paranoid person that I don't want to risk the wrath of the US government for posting these funny pictures of Bush" all the way to the classic standby, "because I want to".

I don't use Freenet, but I also don't simply assume that everybody who searches for perfect anonymity must be a reprehensible criminal.

hmmm

[SinaSa]

Tor - The internet onion!

No, but seriously, the blurb says this is low latency, how that's the case, I fail to see. First client wants to send a HTTP GET or something similar via Tor, so every packet involved needs that info, plus a little bit extra to get it to the next node, plus a little bit more so the end node knows where it needs to be in the end on the return. So that's two extra little bits, then the stuff gets sent one node across which takes its info off and puts new info on.

Where is the low latency here? All this peeling/adding layers to peel off must be fairly time consuming. I'll admit I quite like the idea, and as soon as I click Submit I'm going to download and try it, but I fail to see how this can be faster than say, InvisibleIRC (IIP) was.

Not Like Freenet

[gclef]

Wow. Lots of DefCon related stories.

Anyway, for those asking, no, this isn't quite like Freenet. In TOR, you decide which points you want to send traffic through (and negotiate encryption keys with each one individually), and, unlike FreeNet, you can tunnel existing protocols over it (like, say http).

There's a lot of promise here, but in his talk, he was looking for sites that had at least 1Mbps up & down speeds for nodes. This isn't quite like Peekabooty, in that right now they're not looking for everyone to run a middleman node.

Is the route preselected?

[brett42]

From the couple of days I spent actually working in my highschool cisco class, I remember each router in a path is supposed to be able to optimize the route a packet is sent on by using local information and the packet's final destination. From what I gather from the limited technical details in the article, this protocol would require knowledge of the entire route at the initial node to handle the 'onion layer' encryption.

Is there some way of optimizing a path through a given number of nodes without keeping huge amounts of information about latency on every two nodes, or is this just bouncing the packet around for a while for anonymity and accepting the added latency, plus possibly the time it takes to detect and resend packets when one node in a path suddenly goes dead, making the custom-encrypted packet worthless?