arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2110.00539cs.LGcs.CR
收藏

差分隐私在张量完备化中的应用

Applying Differential Privacy to Tensor Completion

Fudan University(复旦大学);Shanghai Jiao Tong University(上海交通大学)

Zheng Wei, Zhengpin Li, Xiaojun Mao, Jian Wang

摘要:

张量完备化的目的是基于部分观测的张量来填充缺失或未观测的条目。然而,在许多实际情况下,利用观察到的张量往往会引起严重的隐私问题。为了解决这个问题,我们提出了一个坚实的和统一的框架,其中包含几种方法应用差分隐私的两个最广泛使用的张量分解方法:i)CANDECOMP/PARAFAC~(CP)和ii)Tucker分解。对于每种方法,我们都建立了严格的隐私保证,同时评估了隐私准确性的权衡。在合成数据集和真实数据集上的实验表明,我们的建议在确保强隐私保护的同时,实现了高精度的张量完成。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑