Applying Differential Privacy to Tensor Completion
This work addresses privacy concerns in tensor completion for applications handling sensitive data, representing an incremental improvement by adapting existing privacy techniques to tensor methods.
The paper tackles the privacy risks in tensor completion by proposing a framework to apply differential privacy to CP and Tucker decompositions, achieving high accuracy while ensuring strong privacy protections.
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.