BecomingJulie
Bluelight Crew
If you controlled enough internal and exit nodes on the TOR network, and modified the software (which by default runs from RAMdisk, with a minimal kernel having only the necessary drivers and doesn't log anything), you might in theory get lucky decrypting packets that were sent through "your" nodes. And anything downstream of an exit node is "in the clear" (though possibly still encrypted by one layer of SSL, if the underlying protocol is HTTPS) anyway. The exit node does necessarily know the final destination address. (Analogy: The payload may have been delivered in an armoured security van, and the possibly-crooked driver knows where it went even if they were unable to open it and see what was inside; but the delivery company have no power over what the person who signed for it does with it afterward .....)
All that said, it's far-fetched for your entire conversation both ways to go through exactly the right (wrong?) series of hops. They may get one packet -- up to 1500 characters -- if they're very, very lucky. And the final destination address isn't massively secret anyway. Of course, there are servers within the Tor network itself (identified by .onion addresses); in which case there is no exit node to worry about -- just the legitimacy of the person operating the server.
All that said, it's far-fetched for your entire conversation both ways to go through exactly the right (wrong?) series of hops. They may get one packet -- up to 1500 characters -- if they're very, very lucky. And the final destination address isn't massively secret anyway. Of course, there are servers within the Tor network itself (identified by .onion addresses); in which case there is no exit node to worry about -- just the legitimacy of the person operating the server.
