Secure and Efficient Hybrid Data Deduplication in Edge Computing

HyungJune Shin, Dongyoung Koo, Junbeom Hur
2022 ACM Transactions on Internet Technology  
As an extension of cloud computing, edge computing introduces additional intermediate devices, called edge nodes near clients, providing computing services on behalf of the central cloud more efficiently. Although edge computing brings several benefits such as low latency and bandwidth savings on the edge side, rapid increase in the amount of data transmitted to the central cloud hinders efficient utilization of the storage system on the central cloud side especially when the data from edge
more » ... ces are encrypted. To mitigate this issue in a privacy-preserving manner, data deduplication techniques for encrypted data have been extensively studied to enhance both the security and efficiency in the conventional cloud system with two different approaches. A server-side secure deduplication approach protects data privacy, but impairs network efficiency by allowing duplicate uploads; while a client-side one improves network efficiency, but suffers from potential information leakage due to its vulnerability to the side-channel attack. In this paper, we propose a hybrid secure deduplication scheme for edge computing, which guarantees both advantages of the aforementioned two approaches. Specifically, our scheme guarantees data privacy by applying the server-side deduplication technique between the client and the edge nodes, and maximizes network efficiency through the client-side deduplication technique between the edge nodes and the cloud. In addition, we devise a novel additively homomorphic encryption for efficient deduplication operations in the resource-limited edge nodes. Based on our experimental results, the proposed scheme reduces the communication costs by approximately 2.5 times for a storage server when the duplicate ratio is 50%, and the response time is reduced by about 2 times when the data size is 16 MB.
doi:10.1145/3537675 fatcat:b4uhxfhqgjbhrhvc5upyaqp7bq