The present disclosure relates generally to authentication of voice communications. Methods performed by a user device for mutually authenticated communications can include creating a first communication channel with a backend, creating a secure session across a second communication channel with the backend, receiving a first identification message from the backend via the second communication channel, receiving a second identification message from the backend via the first communication channel, sending an attestation that the second identification message matches the first identification message to the backend via the second communication channel, receiving a second step authorization instruction from the backend via the second communication channel, assessing the identity of the user, and delivering an authorization response to the backend via the second communication based of the assessed identity of the user.