About the SRP Protocol: http://en.wikipedia.org/wiki/Secure_remote_password_protocol
I can see that the generation of the session key (K) is perfectly safe, but in the last step the user sends proof of K (M). If the network is insecure and the attacker in the midlle captures M, he would be able to authenticate without having K. right?
A little background
Well known values (established beforehand):
The users password is established as:
The authentication:
The answer to your question:
As you can see, both parties calculate K (=the session key) separately, based upon the values available to each of them.
If Alice's password P entered in Step 2 matches the one she originally used to generate v, then both values of S will match.
The actual session key K is however never send over the wire, only the proof that both parties have successfully calculated the same session key. So a man-in-the middle could resend the proof, but since he does not have the actual session key, he would not be able to do anything with the intercepted data.