Cloud federation paradigm can improve cloud service providers' (CSPs) profits by renting their idle resource to other federation members. However, these CSPs have the risk that they cannot fulfill their scalability commitment when some of their customers have large short-term resource demand. To reduce this risk, we design a reinsurance-emulated collaboration mechanism in a broker-based cloud federation. Reinsurance is an insurance policy which transfers all or part of insurance business in order to scatter the risk to other insurers. Similar to insurance companies, in our proposed model, each CSP determines its resource retention for its future demand. We design an exact method to determine each CSP's retention, with the aim of maximizing its expected profit. Once the CSP's retention cannot meet its future demand, it reduces the risk by outsourcing part of the requests to others. After every CSP determines its retention, the broker will make an assignment to maximize the resource utilization so as to reduce the risk of the CSPs. We designed an algorithm to maximize the resource utilization, with a guarantee of not worse than half of the optimal solution. Simulation results show that our proposed algorithm is efficient.