Authors

Axel W. Krings, Jean-Louis Roch and Samir Jafar

Abstract

This research addresses certification of large distributed applications executing in hostile environments, where tasks or the results they produce may have been corrupted due to benign or malicious act. We extend recent results addressing applications with task dependencies and introduce new probabilistic certification algorithms that establish whether the computations have been massively attacked. The probabilistic approach does not make any assumptions about task behavior as the result of an attack and certification errors are only due to unlucky random choices. Bounds associated with certification are provided for general graphs and for out-trees found in medical image analysis applications of the French agtime project