In the bizantin agreement, node defects are modeled as Byzantine errors. In this error model, faulty nodes must be arbitrary, including a malicious attempt to prevent non-faulty nodes from reaching an agreement. In particular, unusual knots can collaborate with each other. Peter Robinson was partially supported by the Seventh Framework Programme of the European Community (7/2007-2013) under the ASAP project, grant agreement No 619706. A Byzantine agreement is formally defined to meet the criteria of agreement, validity and termination. The agreement simply requires each non-defective player to emit the same bit. Validity excludes the trivial solution of always emitting a particular bit by requiring that the value of the agreement be proposed at least once. The deadline means that each minute is necessary to be finalised. The bizantin agreement is formally reached if our main contribution is a circular algorithm (O (log ^3 n)), which reaches the choice of the Byzantine guide in the presence of (O (O ^^^1/2 – varepsilon }) of the Byzantine nodes (for a small constant (epsilon > 0)) and a leak to (O (sqrt{n}/text {polylog} }) per turn (n being the stable size of the network). The algorithm chooses a leader with a probability of at least (1-n^varOmega (1)} ) and guarantees that it is an honest node with a probability of at least (1-n ^-varOmega (1)} ); Assuming the algorithm is successful, the identity of the leader is known to a fraction (1-o(1)) of the honest nodes. Our algorithm is fully distributed, easy and easy to implement.

It is also scalable because it is run in polylogarithmic time and nodes must send and receive polylogarithmic size messages only per turn. To our knowledge, our algorithm is the first scalable solution to choose the Byzantine guide in a dynamic network with a high rate of emigration; Our protocol can also be used to resolve Byzantine agreements in a simple way. We also show how to implement a public piece (almost everywhere) with constant distortion in a dynamic network with Byzantine nodes and provide a mechanism that allows honest nodes to reliably store information on the network that could be of independent interest. . . .