It doesn’t. It doesn’t even try. It’s not necessary to verify the identity of the other party to establish a secure call. In a normal PSTN phone call, what happens if you call someone’s number, and his wife answers the phone? Do you sound a klaxon horn and blow a fuse? No. You use your brain to figure it out. That’s just how phones work, and it’s no big deal. It’s certainly no reason to fail to make a secure connection. The most important wiretapping vulnerability is a Man-in-the-Middle (MiTM) attack, which ZRTP guards against by using either a Short Authentication String, or key continuity, or both.
Of course, it helps if you know the identity of the caller before you answer the phone, like the Caller ID in the PSTN. The SIP protocol attempts to address that problem in the signaling . It is a different problem, and certainly worthy of attention, but it is not the job of ZRTP. ZRTP does not begin until after the user answers the phone and the call is underway. ZRTP merely establishes a secure wiretap-resistant connection to another ZRTP endpoint, and does it very well by narrowing the scope of its mission.
I don’t know why so many people get hung up on this question of making an “authenticated phone call”. It’s a hard problem, and not worth the effort, in my opinion. Most phones, both at home and at work, are used by more than one person. And many people use other people’s phones on a regular basis. There is not a one-to-one relationship between people and phones. Then there is the problem of establishing a digital identity. We could have a complex bureaucracy create a public key infrastructure that issues a certificate that we can attach to my phone, which can be displayed by your phone. Not only is that of questionable value in my opinion, but it’s also hard. A number of clever people, including my friend Carl Ellison, have written about the complexity of creating unambiguous unique names and attaching them to people.
It’s a mistake to view the world through a radar screen-- you must also use your eyes. And your common sense. The ZRTP protocol cannot tell you the name on the birth certificate of the person you are talking to. Or that the person you are talking to is telling the truth. And it cannot tell you if the other “endpoint” is then forwarding the call to another device. But neither can anything else.