SA-LES-Dedup: Similarity-Aware Lightweight Encrypted Block Deduplication for Cloud–Edge Storage

Abstract

Encrypted deduplication improves storage efficiency while preserving data confidentiality, but cloud-edge deployments still face low-entropy guessing, excessive cloud traffic, and costly edge-index lookups. This paper presents SA-LES-Dedup, which combines OPRF-assisted block-key derivation, edge-local/cloud-global deduplication, lightweight proof of ownership, and SimHash-based candidate selection. Clients derive deterministic block keys and exact tags from OPRF outputs. Edge nodes first search a bounded candidate set and then fall back to the complete local index before forwarding unresolved blocks to the cloud. Similarity tags only accelerate lookup, while exact tags and ownership proofs determine deduplication. We analyze leakage, correctness, and security under honest-but-curious edge and cloud servers. Model-based results show that cloud-bound ciphertext traffic decreases with the edge-local duplicate ratio, while candidate-first lookup reduces edge-index cost when similarity concentration is high. The scheme remains correct under similarity false positives and false negatives.