I've read that obfs4 and scramblesuit are very resistant ("immune" is so optimistic) to such things as active probes performed by the Great Firewall, which can quickly probe and detect older transports (and of course vanilla ORports), plus the older transports and ORports are subject to relatively quick detection through deep packet inspection once a user connects from there.
Does it make sense to offer older more vulnerable transports along with newer more secure ones? If my bridge offers both obfs3 and obfs4, does that just mean that as soon as someone in China uses obfs3 it's detected and my IP address is blocked, making the obfs4 port unusable from there as well even though it would have avoided detection on its own? More fundamentally, does the bridge address server also publish vanilla ORports for those bridges which offer obfs4, and does a Chinese user accessing my bridge's ORport doom my entire bridge to immediate blockage from there?
I can't imagine the GFW would be so kind as to only block the ORport's specific port number, I assume it blocks the entire bridge IP address, making all transports useless if any single one of them is detected. Would it be better to only offer obfs4 to avoid detection and blockage via older transports?
On Sun, Jan 17, 2016 at 11:34:48AM -0800, Rick Huebner wrote:
I can't imagine the GFW would be so kind as to only block the ORport's specific port number, I assume it blocks the entire bridge IP address, making all transports useless if any single one of them is detected. Would it be better to only offer obfs4 to avoid detection and blockage via older transports?
So far, the GFW blocks the bridges it discovers by IP:port. Therefore, you can run two bridges on your machine and if one gets discovered, the other one should still be reachable. Of course, that could change any moment.
Ideally, we would like bridges to only run modern transports such as obfs4. Unfortunately, the following bug is still in the way, requiring the vanilla OR port to be reachable: https://bugs.torproject.org/7349
Cheers, Philipp
So far, the GFW blocks the bridges it discovers by IP:port. Therefore, you can run two bridges on your machine and if one gets discovered, the other one should still be reachable. Of course, that could change any moment.
Wow, that's a pretty huge and inexplicable oversight on their part. Hard to believe they crippled their efforts so badly by missing something so obvious, but I can't think of any reason they'd have to restrict themselves so generously on purpose. Maybe a conscripted engineer's subtle resistance sabotage? OK, whatever, hopefully it gives time for the vulnerability to be fixed before they exploit it.
tor-relays@lists.torproject.org