How to Expose an Eavesdropper (1984)
fermatslibrary.com
fermatslibrary.com
The interlock technique is very clever, I have never seen anything quite like it. However, at first glance, I do not know where the data blocks MA and MB for the interlocking come from. If they are hard-coded, breaking the scheme is trivial for C. If they are dependent on A or B, then we presuppose knowledge about the other party, and in that case, why not just use public key infrastructure with public keys for A and B?
I think the most important stepping stone that makes it hard to apply this protocol in practice is that you somehow need these blocks MA and MB of information that C does not have. If we are talking about voice samples here, it is likely that you can recognize them, but not your computer. Therefore, you'd also need some complex UI interactions to make sure it really is the other person. I'm having a hard time thinking about a practical application and its security implications right now.
An eavesdropper cannot then substitute the contents of either MA or MB without either breaking the decoding process or dropping portions of the conversation. Either way, the eavesdropper cannot fully conceal their presence.
[0] Bellovin, Steven M., and Michael Merritt. "An attack on the interlock protocol when used for authentication." IEEE Transactions on Information Theory 40.1 (1994): 273-275.
PDF at http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.112...
From the abstract,
> [...] We demonstrate that an active attacker can, at the cost of a timeout alarm, bypass the password exchange, and capture the passwords used. Furthermore, if the attack is from a terminal or workstation attempting to contact a computer, the attacker will have access before any alarm can be sounded.