> From reading on OnionCat , the clients are essentially hidden services
> once a connection is made it is bidirectional.
No, OC is just a daemon shuffling data back and forth across a
Tor HiddenServicePort. Tor provides a bidir return path to the
source, which the listener (OC) can use, if it thinks it should...
> If A initiates a connection
> to B , A can be sure he/she is talking to B
Yes, up to the 80-bit addressing of Tor. OC translates your request
for a v6 address into an onion address and puts that stream through
Tor.
> but the opposite isnt true .So
> if B has to sure he/she is indeed talking to A , he/she has to initiate a
> connection to A [..... to query and confirm it .....].
Yes. Because B's onion is seeing no onion source address. And B's
OC is seeing an arbitrary v6 source address.
Since most protocols require a reverse channel, it's actually B that
is more at risk of sending their data off to onions unknown. Luckily,
that is where B (if human and not a dolt) usually notices something
is broken and quits it.
And it's kind of pointless to do such spoofing because if A wanted
B's return stream, it should have just asked for it. So it would just be
for the lol's of A blindly convincing B (or B's computer, app, etc) to
disclose something to C.
> Which is what torchat does to authenticate both the parties
> , even if OnionCat is being used the same has to be done to ensure both the
> people know who they are talking to. Am I right in my observation ??
Yes, and as before, OC had plans to do a little OCtoOC ping pong too.
If running IPSEC, etc over v6, learning or making stashes of source
key-v6 associations, that might do it too, more work, same thing.