Applying Differential Privacy to Tensor Completion

2021/10/01 by Zheng Wei, Zhengpin Li, Wei, Zheng +5
Computer Science · Mathematics · Medicine · #Advanced Neuroimaging Techniques and Applications #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Tensor decomposition and applications #cs.CR #cs.LG

paper · pdf · doi:10.48550/arxiv.2110.00539

We have fixed some format issues in the previous version. 17 pages, 4 figures

openalex publication_date 2021/10/01 · arxiv created 2022/02/14 · arxiv updated 2022/02/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Tensor completion aims at filling the missing or unobserved entries based on partially observed tensors. However, utilization of the observed tensors often raises serious privacy concerns in many practical scenarios. To address this issue, we propose a solid and unified framework that contains several approaches for applying differential privacy to the two most widely used tensor decomposition methods: i) CANDECOMP/PARAFAC~(CP) and ii) Tucker decompositions. For each approach, we establish a rigorous privacy guarantee and meanwhile evaluate the privacy-accuracy trade-off. Experiments on synthetic and real-world datasets demonstrate that our proposal achieves high accuracy for tensor completion while ensuring strong privacy protections.

Related

• Local Differential Privacy for Tensors in Distributed Computing Systems
• Online and Differentially-Private Tensor Decomposition
• Synthetic Data Generation and Differential Privacy using Tensor Networks' Matrix Product States (MPS)
• Protecting Big Data Privacy Using Randomized Tensor Network Decomposition and Dispersed Tensor Computation
• Deep Learning with Gaussian Differential Privacy