Leveraging Network Functions Virtualization Orchestrators to Achieve Software-Defined Access Control in the Clouds

摘要

Network Functions Virtualization (NFV) has been widely recognized as an effective way to implement and consolidate hardware-based network functions by using software-based approaches, with a potential to significantly reducing CAPEX and OPEX. In particular, NFV orchestrators (e.g., Tacker, Cloudify, and ONAP) play a vital role in managing and orchestrating various virtualized network resources (e.g., VMs, Virtualized Network Functions), and TOSCA is one of the standard data models to fulfil such a role. However, it remains unclear how the security mechanisms can be seamlessly integrated into the entire lifecycle of those virtualized network assets. Starting with a comparative analysis on the available NFV orchestrators, we extend the TOSCA model to incorporate security attributes of interest, and leverage the extended model to create access control policies at cloud scale. Specifically, a security orchestrator is developed, which contains a TOSCA-parser and a novel tenant-specific access control paradigm. One of the salient features of our security orchestrator is that it allows to dynamically generate access control models and policies for different tenant domains, resulting in a flexible and scalable protection coverage that is across different NFV layers and multiple data centers. To validate its feasibility and effectiveness, we develop a security orchestrator prototype and test its performance with respect to throughput, scalability, and adaptability. The experimental results demonstrate that all the desirable properties can be achieved, and the throughput of our security orchestrator can be maintained at a satisfactory level regardless of the varying number of tenants, users, or objects that are deployed in the cloud.