In the secure communication problem, we focus on safe termination. In applications such as electronic transactions, we want each party to be ensured that both sides agree on the same state: success or failure. This problem is equivalent to the well-known coordinated attack problem. Solutions exist. They, however, concentrate on the probability of disagreement, and attack incentives have been overlooked so far. Furthermore, they focus on a notion of round and are not optimal in terms of communication complexity. To solve the safe termination problem, we revisit the Keep-in-Touch protocol that we introduced in 2003. Considering the communication complexity, the probability of unsafe termination, and the attack incentive, we prove that the Keep-in-Touch protocol is optimal. (c) 2006 Elsevier B.V. All rights reserved.