As we saw in Section 2.6, a QKG system with a limited lifetime can only generate a limited amount of shared secret key and can be replaced with a pregenerated courier delivered shared key with increased security. Some of the messages of each round need to be authenticated and once the initial key is used a key generated in previous rounds needs to be used. Section 1.2 explains why Eve will have a small but non-zero amount of information about that key.
In Chapter 6 the method of encrypting the authentication tags proposed by Wegman and Carter is shown to be unsuitable when completely secret keys are not available. Using that method in a QKG system would each round give Eve more knowledge about the hash function, which would limit the lifetime of the system to the time when she is expected to know enough to launch an attack.
Chapter 4 gave many examples of the dangers of good security only on average. Chapter 6 reveals that even if Eve initially has very little information about the authentication key, when she has seen Alice's message/tag pair only the expectation value of her knowledge is bounded. If we assume that the authentication tag is sent along with the message and that Eve only needs to forge one message to gain enough power over the key growing process to be able to forge the next message, the risk that Eve is in control of the QKG system will increase for each round and approach unity without Eve ever risking being detected. This would put a theoretical limit to the lifetime of the system.
Fortunately simple solutions exist and two of them are also presented in Chapter 6. They both force Eve to make her attack before she knows that it will succeed by making sure Alice will not send the authentication tag until either Bob has received the message or Eve has done something that would reveal her if she cannot produce the correct tag for her forged message. A QKG system might already have similar properties since a round normally consists of a dialogue of several messages and an authentication tag for all of them at the very end of the round. Whether that is enough to keep the system secure depends on the details of the system, but implementing one of the solutions is cheap and requires no deep analysis of the system.
Finally note that the proposed solutions only makes the authentication secure if Eve's initial knowledge of the key is limited. If the privacy amplification only limits her average knowledge of the authentication key, she will eventually know enough of the key to safely launch an attack regardless of the details of the authentication process.