Secure and Lightweight Deduplicated Storage via Shielded Deduplication-Before-Encryption

Outsourced storage should fulfill confidentiality and storage efficiency for large-scale data management. Conventional approaches often combine encryption and deduplication based on deduplication-after-encryption (DaE), which first performs encryption followed by deduplication on encrypted data. We argue that DaE has fundamental limitations that lead to various drawbacks in performance, storage savings, and security in secure deduplication systems. In this paper, we study an unexplored paradigm called deduplication-before-encryption (DbE), which first performs deduplication and encrypts only non-duplicate data. DbE has the benefits of mitigating the performance and storage penalties caused by the management of duplicate data, but its deduplication process is no longer protected by encryption. To this end, we design DEBE, a shielded DbE-based deduplicated storage system that protects deduplication via Intel SGX. DEBE builds on frequency-based deduplication that first removes duplicates of frequent data in a space-constrained SGX enclave and then removes all remaining duplicates outside the enclave. Experiments show that DEBE outperforms state-of-the-art DaE approaches.