I recently discovered this project and it looks very interesting. I understand why broadly distributed data is effectively protected against censorship, DDoS attacks, since there is no single point of failure and anyone can pin and redistribute it. However, it also seems that it takes some time to propagate any new information. So the question is, if the attacker targets the publisher by its identity, wouldn’t it be much easier to “silence” the data which hasn’t been propagated yet? And what are possible counter-measures if this is ever an issue?
Ideally, use something like Tor and ephemeral peer IDs; basically, make your node a moving, hidden target. Unfortunately, we currently don’t have tor support (although we’re working on it) and don’t have support for ephemeral IDs either.
For anyone interested in using IPFS over Tor, it seems that this issue is being tracked here: https://github.com/ipfs/notes/issues/37
From the latest post on Aug 14, it sounds like there might be a professional audit of the IPFS stack underway (?) or planned – which seems to be one of the last things before Tor support is made public.