3 ms·
Wait, what? The paper says that NetFlow is not optimized for this kind of attack, that it is merely representative of flow data available from core routers in
by schoen 12y ago
Wait, what?
The paper says that NetFlow is not optimized for this kind of attack, that it is merely representative of flow data available from core routers in general ("traffic monitoring functionality built into the routers of major IXs and ASs, such as Cisco’s NetFlow"), and that you have to do research to figure out how to make it work. The authors aren't saying that Cisco intended to facilitate traffic analysis attacks against users, or that using Cisco routers in a network you set up makes you or your users more vulnerable to traffic analysis attacks by other people.
I'd love to see much more research on new approaches to anonymity, but it's also important to understand the tradeoffs. In the face of active attacks like this one, it seems that a network needs to be synchronously padded (all participants and potential participants always transmit and receive at a fixed speed, regardless of whether they're communicating or not) or else high-latency (you wait a long time to forward communications far enough through the network that they genuinely could have been from anyone in a large population based on detailed traffic statistics). Even the padded versions might be vulnerable to an adversary who can actually disrupt or degrade people's connections (to prevent particular nodes from sending or receiving the amounts of traffic that they're expected to by the protocol). Even a very successful Kickstarter project isn't going to provide a path to escape these tradeoffs.