Paging and exit overheads have been proven to be the performance bottlenecks when adopting Searchable Symmetric Encryption (SSE) with trusted hardware such as Intel SGX for keyword search. This problem becomes more serious when incorporating ORAM and SGX to design oblivious SSE schemes such as POSUP [1] and Oblidb [2] which can defend against inference attacks. The main reason comes from high round communication complexity of ORAM and constrained trusted memory created by SGX. To overcome this performance bottleneck, we propose a set of novel SSE constructions with realistic security/performance trade-offs. Our core idea is to encode the keyword-identifier pairs into a bloom filter to reduce the number of ORAM operations during the search procedure. Specifically, Construction 1 loads the bloom filter into the enclave sequentially, which outperforms about$1.7\times$when the dataset is large compared with the performance of the baseline that directly combines ORAM and SGX. To further improve the performance of Construction 1, Construction 2 classifies keywords into groups and stores these groups in different bloom filters. By additionally leaking the keywords in search token belonging to which groups, Construction 2 outperforms Construction 1 by$16.5\sim 36.8\times$and provides an improvement of at least one order over state-of-the-art oblivious protocols.

Digital twin (DT) constructs virtual counterparts of physical devices to monitor and optimize their life cycle processes. With the emergence of industry 4.0, Industrial Internet of Things (IIoT) has became the backbone of the DT by providing a fundamental way to transform physical devices to their virtual counterparts. With the deployment of IIoT, built-in sensors enable real-time collection of critical DT data involving various physical parameters associated with devices during their life cycle. However, traditional data sharing services rely on a centralized infrastructure, which inevitably brings severe security threats to share large volume of sensitive DT data derived from numerous sensors. To address the above issue, this paper presents a blockchain based Multi-users Oblivious Data Sharing scheme (MODS) for the digital twin system in the context of IIoT. MODS supports a broad range of security properties including confidentiality, obliviousness, and access control for the DT data stored on the blockchain. MODS adopts a hybrid design approach by combing trusted hardware and cryptography to achieve well balances between security and efficiency. To demonstrate the design advantages of MODS, we explore the design space of a multi-users oblivious data sharing scheme by using pure cryptographic approach, which incurs several design tradeoffs that must be addressed. We show that MODS performs well in these tradeoffs. A comprehensive evaluation has been conducted to demonstrate that MODS is practical to support secure data sharing via blockchain for IIoT.