Distributed integrity checking for systems with replicated data

This work presents a new comparison-based diagnosis model and a new algorithm, called Hi-Dif, based on this model. The algorithm is used for checking the integrity of systems with replicated data, for instance, detecting unauthorized Web page modifications. Fault-free nodes running Hi-Dif send a task to two other nodes and the task results are compared. If the comparison produces a match, the two nodes are classified in the same set. On the other hand, if the comparison results in a mismatch, the two nodes are classified in different sets, according to their task results. One of the sets always contains all fault-free nodes. One fundamental difference of the proposed model to previously published models is that the new model allows the task outputs of two faulty nodes to be equal to each other. Considering a system of N nodes, we prove that the algorithm has latency equal to log₂N testing rounds in the worst case; that the maximum number of tests required is O(N²); and, that the algorithm is (N-1)-diagnosable. Experimental results obtained by simulation and by the execution of a tool implemented applied to the Web are presented.