The shared memory model matches important classes of systems deployed over dynamic networks, as for example, fault-tolerant and high available data centric services. Consensus is a fundamental building block able to realize such reliable distributed systems. Unlike the classical setting where the full set of participants and their identities are known to every process, dynamic networks preclude such global knowledge to be available. In this paper, we investigate and present protocols to solve fault-tolerant consensus in an environment with unknown participants that communicate via shared memory.