A large of smart nodes is connected in Internet of Things (IoT). These nodes in IoT often are resource-constrained nodes, so they have a greater need to collaborate with other nodes for performing advanced service and applications. When there are many providers who can provide similar collaboration service, a collaboration requestor (CR) node will have to make the decision of the “best” collaboration partner (CP). Many trust based approaches have been proposed to determine the “best” CP. However, there are limitations in the existing approaches. The first limitation is ignoring the cross-domain collaboration interactions exist among nodes. The second limitation is only considering a single objective formulation in which, the most trustworthy node are selected as the “best” CP. However, in fact the most trustworthy node does not always bring the “best” collaboration benefit. This paper proposes a multi-objective cross-domain collaboration management approach called MCCM in which collaboration benefit of CR node and trust value of CP node are considered as two optimization objects. We analyze MCCM performance in terms of trust accuracy, convergence, and resiliency against attacks. In addition, we also compare MCCM with trust based methods. Experiment results demonstrate that MCCM outperforms trust based collaboration partner selecting approaches in terms of collaboration partner prediction accuracy against recommendation attacks.

物联网（IoT）中连接了许多智能节点。物联网中的这些节点通常是资源受限的节点，因此他们更需要与其他节点协作以执行高级服务和应用程序。当有许多提供商可以提供类似的协作服务时，协作请求者（CR）节点将必须做出“最佳”协作伙伴（CP）的决定。已经提出了许多基于信任的方法来确定“最佳” CP。但是，现有方法存在局限性。第一个限制是忽略节点之间存在的跨域协作交互。第二个限制仅考虑单个目标公式，其中最值得信赖的节点被选为“最佳” CP。然而，实际上，最值得信赖的节点并不总是带来“最佳”协作收益。本文提出了一种多目标跨域协作管理方法，称为MCCM，其中将CR节点的协作收益和CP节点的信任值视为两个优化对象。我们从信任准确度，收敛性和抵御攻击能力方面分析MCCM性能。此外，我们还将MCCM与基于信任的方法进行了比较。实验结果表明，在针对推荐攻击的协作伙伴预测准确性方面，MCCM优于基于信任的协作伙伴选择方法。本文提出了一种多目标跨域协作管理方法，称为MCCM，其中将CR节点的协作收益和CP节点的信任值视为两个优化对象。我们从信任准确度，收敛性和抵御攻击能力方面分析MCCM性能。此外，我们还将MCCM与基于信任的方法进行了比较。实验结果表明，在针对推荐攻击的协作伙伴预测准确性方面，MCCM优于基于信任的协作伙伴选择方法。本文提出了一种多目标跨域协作管理方法，称为MCCM，其中将CR节点的协作收益和CP节点的信任值视为两个优化对象。我们从信任准确度，收敛性和抵御攻击能力方面分析MCCM性能。此外，我们还将MCCM与基于信任的方法进行了比较。实验结果表明，在针对推荐攻击的协作伙伴预测准确性方面，MCCM优于基于信任的协作伙伴选择方法。