DMTL: An Adaptive Integrity Verification Scheme for Dynamic Cloud Datasets

Abstract

With the widespread application of cloud computing, large-scale datasets in fields such as deep learning are increasingly stored in the cloud. Advanced techniques for data integrity verification are necessitated due to the frequent incremental updates of these datasets. In practice, existing Merkle tree-based schemes face challenges, including high computational costs, low real-time performance, and inefficient handling of incremental updates. To address these issues, a novel data structure named dynamic Merkle tree ladder (DMTL) was proposed in this paper, which enhanced Merkle trees by establishing ladder rungs for each dataset and incorporating a flexible dataset partition strategy. Based on the DMTL, we designed an integrity verification scheme that supported adaptive incremental updates of cloud datasets. Experimental results demonstrated that our scheme had outperformed mainstream schemes in dynamic operation efficiency, especially under workloads with intensive data insertions. Security analysis further showed that our scheme could defend against malicious behaviors effectively.