Replicating data changes through distributed invalidation

A computer-implemented method for replicating data changes through distributed invalidation includes receiving, by a distributed database system, an instruction to change a data element in a table. The distributed database system includes at least a first server and a second server. A first copy of the table is stored on the first server, and a second copy of the table is stored on the second server. The method further includes in response to the instruction, determining that the data element is secured by a replication key that is stored on a shared key management system that is accessible by the first server and by the second server, wherein the replication key is unique to the data element. The method further includes invalidating the replication key and modifying the first copy of the table on the first server according to the instruction that is received.