Verifiable and Secure SVM Classification for Cloud-Based Health Monitoring Services

In cloud-based health monitoring services, support vector machine (SVM) classification techniques are often utilized by medical institutes to build medical decision models, which can be outsourced to a cloud server for producing medical decisions based on medical features from remote clients. In this article, we propose a verifiable and secure SVM classification scheme (
$\mathsf {VSSVMC}$
) for cloud-based health monitoring services in a malicious setting, where the cloud server may return invalid decisions. By constructing verifiable indices,
$\mathsf {VSSVMC}$
ensures the verifiability of medical decisions, which enables clients to detect whether the cloud server returns incorrect or incomplete medical decisions. Symmetric key encryption is leveraged to ensure the confidentiality of the medical decision model and medical data with computational efficiency. We give security and verifiability definitions and provide formal security and verifiability proofs for
$\mathsf {VSSVMC}$
. Performance analyses show that
$\mathsf {VSSVMC}$
is extremely efficient in terms of computation, communication, and storage. Experimental evaluations demonstrate that
$\mathsf {VSSVMC}$
achieves microsecond-level execution time with kilobyte-level communication and storage overheads on the tested data set.