Some German guy got got for logging into IRC via encrypted wifi, the cops did some war driving and correlated timing of traffic spikes with IRC messages until they had a profile with better hit probability than a DNA match.
The best thing about that? They didn’t even need a search warrant as our genius was broadcasting the side-channel to the whole neighbourhood.
Sounds interesting, got any links for further reading on that?
I can’t quite connect the dots between wifi/internet traffic spikes when IRC is so light on traffic that it’s basically background noise and war driving.
When you send a message, that usually fits into an IP packet. That gets completely encrypted by the wifi, but you know that a data packet approximately that size has been sent at exactly that time. Simultaneously, you watch the IRC channel and see when messages are arriving from your suspect, or someone else types a message and that should correlate with another encrypted wifi package.
The mistake was a) using wifi, exposing the data in the first place and b) not torrenting while you’re chatting. That would’ve obscured the time correlations.
I have an understanding of the underlying concepts. I’m mostly interested in the war driving. War driving, at least in my understanding, implies that someone, a state agency in this case, physically went to the very specific location of the suspect, penetrated their (wireless) network and therefore executed a successful traffic correlation attack.
I’m interested in how they got their suspects narrowed down that drastically in the first place. Traffic correlation attacks, at least in my experience, usually happen in a WAN context, not LAN, for example with the help of ISPs.
Laymen with no understanding here. Obviously there were other mistakes, all of which make sense to me on a rudimentary level, but the first mistake you listed was him using wifi? What is the more secure alternative? Or do you just mean sending data directly over a true wifi connection and not using TOR or another medium?
Timing attacks work, but if they’re running those then they have a pretty good idea as to both sides of the convo.
Put another way, if they’ve got to that point your opsec has already failed.
Well without that they would’ve needed probably months correlating things like “goes to bed at X o clock” with those chat logs. For a whole neighbourhood to then get a search warrant with that. Which of course is not above the capabilities of a state actor but depending on how big a fish he was they might not have bothered spending the resources. Being able to pin-point a house in maybe a day when all you have is a municipality is a whole different ballpark.
