2025
Efficient and Verifiable Privacy-Preserving Convolutional Computation for CNN Inference with Untrusted Clouds
Jinyu Lu, Xinrong Sun, Yunting Tao, Tong Ji, Fanyu Kong*, Guoqiang Yang (* equal contribution)
International Conference On Intelligent Computing (ICIC). 2025 [CCF C]
We propose a novel verifiable privacy-preserving scheme tailored for CNN convolutional layers, which achieves speedups ranging $26\times$ ~ $87\times$ compared to the original plaintext model while maintaining accuracy.
Efficient and Verifiable Privacy-Preserving Convolutional Computation for CNN Inference with Untrusted Clouds
Jinyu Lu, Xinrong Sun, Yunting Tao, Tong Ji, Fanyu Kong*, Guoqiang Yang (* equal contribution)
International Conference On Intelligent Computing (ICIC). 2025 [CCF C]
We propose a novel verifiable privacy-preserving scheme tailored for CNN convolutional layers, which achieves speedups ranging $26\times$ ~ $87\times$ compared to the original plaintext model while maintaining accuracy.
2024
Secure Distributed Document Representation Outsourcing Scheme for Natural Language Processing in Cloud Computing
Jinyu Lu, Xinrong Sun, Yunting Tao, Fanyu Kong*, Chunpeng Ge, Hanlin Zhang (* equal contribution)
IEEE Transactions on Cloud Computing (TCC).Under review. 2024 [CCF B]
We propose a secure and verifiable distributed document representation outsourcing scheme based on the FBoW model in cloud computing. This scheme employs a novel blind method based on orthogonal symmetric matrices to preserve privacy and minimize computational overhead, achieving a 95.42% improvement in performance.
Secure Distributed Document Representation Outsourcing Scheme for Natural Language Processing in Cloud Computing
Jinyu Lu, Xinrong Sun, Yunting Tao, Fanyu Kong*, Chunpeng Ge, Hanlin Zhang (* equal contribution)
IEEE Transactions on Cloud Computing (TCC).Under review. 2024 [CCF B]
We propose a secure and verifiable distributed document representation outsourcing scheme based on the FBoW model in cloud computing. This scheme employs a novel blind method based on orthogonal symmetric matrices to preserve privacy and minimize computational overhead, achieving a 95.42% improvement in performance.