On the Resilience of [Distributed] EAs against Cheaters in Global Computing Platforms

Global Computing (GC) platforms such as BOINC are nowadays considered as the most powerful distributed computing systems worldwide. Based on volunteer computing and various forms of incentives, such architecture also attract cheaters who seek to obtain rewards with little or no contribution to the system. Cheating can be modelized as alteration of output values produced by some or all tasks of the program being executed. In complement of generic approaches typically based on task duplication, this article focus on Algorithm-Based Fault Tolerance (ABFT) technics where the fault tolerance scheme is tailored to the algorithm performed to make it resilient to a limited number of faults i.e. falsified results produced by cheaters in this case. More precisely, the aspects of distributed Evolutionary Algorithm (dEA) against cheaters in environment is studied. Our main contribution consists in the formal analysis of the impact of cheating faults on this context, together with a theoretical proof of convergence towards valid solutions despite the presence of malicious acts. Whereas the resilience of against a simpler model of fault (i.e. crash faults) has been observed in the literature, this study is the first to propose a theoretical proof of this behaviour, additionally against a more complex kind of fault. By the variety of problems addressed by EAs, this study will hopefully promote their usage in the future developments around GC platforms.