The Multi-Agent Patrolling task constitutes a challenging is- sue for the MAS field and has the potential to cover a variety of domains ranging from agent-based simulations to distributed system design. Several techniques have been proposed in the last few years to address the basic multi-agent patrolling task. Recently, a variation of this task was proposed, in which agents can enter or leave the task at will: the patrolling task with an open system setting. A few centralized strategies were also described to address this new problem. In this article, we propose to adapt to a dynamic population a completely decentralized strategy that was proposed for the original basic patrolling task: an auction-based strategy in which agents trade the nodes they have to visit. We describe and compare several entry and exit algorithms on various graph topologies and show the interest of basing these mechanisms on geographical proximity. Finally, we compare this strategy to the centralized ones on simulations with multiple variations in the population of agents, and show that it provides a strong stability and reactivity to changes in the population of agents.