Bypassing the Telegram authentication protocol [pdf]
cert.inteco.es
cert.inteco.es
[1] The Old New Thing has a slew of similar "exploits" and well-explained comments on why they're false. As they aren't properly tagged, I coulnd't find a nice link -- but the blog is fun to browse and read: http://blogs.msdn.com/b/oldnewthing/
An attacker could simply take the source code, change the server IP and public-key and has a working MITM, if he can somehow make the user install this one instead of the original.
A modified client can log, leak or change any message it wants once it has the plaintext. You shouldn't need to understand the protocol at all.
Money quote:
What should an attacker do to launch the attack?
Briefly, he would have to change Telegram’s public key (and the IP address,
or perform a “live” spoofing attack), included in the client, and distribute
the malicious client to the victim(s).
For comparison, this would be equivalent to editing entries in ~/.ssh/known_hosts of the victim and then make them use ssh to your server for a MITM attack. Or rather, if you get someone to use a manipulated SSH client, they are vulnerable. Would you consider this a flaw in the SSH authentication protocol?That's from the same domain as planting a fake CA cert on the client and then proxying its SSL connections.