Cerberus: Privacy-Preserving Computation in Edge Computing

Abstract

Edge computing reduces the overhead of data centers and improves the efficiency of data processing. However, traditional cloud data protection mechanisms are no longer applicable to edge devices. Data leakage and other privacy issues may occur when computation is outsourced to edge nodes. The decentralization raises new privacy challenge for data control, storage and computation. In this work, we present Cerberus, a brand-new framework that preserves data privacy in edge computing by combining blockchain, distributed data storage and trusted execution environment (TEE). Blockchain is used to maintain a global computation state, and also acts as a medium of information interaction. Distributed data storage provides a secure and large-capacity storage. TEE-based off-chain computation guarantees confidentiality and efficiency of data processing. We also implement a prototype of Cerberus using Hyperledger Fabric and Intel SGX. Our evaluation on a sample of data sorting application shows that Cerberus achieves significant speed ups over previous cryptographic schemes. Compared with non secure computation, Cerberus can preserve data privacy without incurring much performance loss.