As of a week ago, I now find myself living in one of the more oppressively surveilled states in the world. I do not like this on ideological grounds. Therefore, I will acquire a vpn subscription for my family's traffic online, and will set up a default route for the building to the supplier (probably crpytostorm.is), via a dd-wrt router. Being an honest citizen who does nothing wrong online (and I assume the same for the rest of the house), my "threat model" is minimal, and therefore one might naïvely assume that this would be enough to prevent the Food Standards Agency from looking at my browsing history.
However, I recently read about how private -- or not -- vpns might be; correlation attacks are highly likely to be able to defeat such measures in an internet where essentially everything is logged.
Therefore, I'd like to ask this question: if one were to maximally use your available bandwidth continuously, constantly maintaining an encrypted tunnel to a VPN provider (who we assume now is a good faith actor), would that be sufficient to mitigate significantly against correlation attacks? It seems like that would be a "relatively" trivial (albeit annoying) task for a router to perform: estimate the current/near-term-future available bandwidth, send a large number of, e.g. TCP NOP packets at a rate designed to saturate the connection, and adjust their volume/frequency to not disrupt other data.
I have little doubt that my ISP wouldn't like me actually using the bandwidth I'd paid for, but beyond that, it seems like it would make it at the least marginally harder to perform correlation attacks on an encrypted tunnel. I imagine that the moderately beefy VPN server I'd connect to could silently drop the NOP packets (or equivalent) with a minimal performance hit.
Am I right? This seems like an obvious idea -- what is it usually called? Would this method work to mitigate against correlation attacks? Is there an open-source implementation in, e.g., dd-wrt to perform it?