Mining large data requires intensive computing resource and data mining expertise, which might be unavailable for many users With widely available cloud computing resources, data mining tasks ca now be moved to the cloud or outsourced to third parties to save costs In this new paradigm, data and model confidentiality becomes the majo concern to the data owner. Data owners have to understand the potentia trade-offs among client-side costs, model quality, and confidentialit to justify outsourcing solutions. In this paper, we propose the RASPBoos framework to address these problems in confidential cloud-base learning. The RASP-Boost approach works with our previous develope Random Space Data Perturbation (RASP) method to protect data confidentialit and uses the boosting framework to overcome the difficulty o learning high-quality classifiers from RASP perturbed data. We develo several cloud-client collaborative boosting algorithms. These algorithm require low client-side computation and communication costs. The clien does not need to stay online in the process of learning models. We hav thoroughly studied the confidentiality of data, model, and learning proces under a practical security model. Experiments on public dataset show that the RASP-Boost approach can provide high-quality classifiers while preserving high data and model confidentiality and requiring lo client-side costs.