使用者的資料隱私在雲端機器學習中是一個非常重要的議題，在本篇論文中 我們提出一種新的威脅來自上級權威機構，權威機構可以要求使用者與雲端服務 供應商給出隱私資料，權威機構也可以監控使用者在雲端服務的行為，我們提出 了一種具有可否認性的可學習圖像加密的保護方法，當使用者遭到了上級權威機 構脅迫交出在雲端服務平台中的訓練資料，可以生成出一把假的密鑰使脅迫者進 而去解密出假的資料，保護使用者在雲端服務平台中的隱私資料，而雲端服務供 應商也因為圖像被加密無法得知使用者的隱私資料。我們使用了分佈式多模型預 測查詢提升預測準確率.因可學習圖像加密準確度降低的問題。我們也將我們的 方案與其他可學習的加密技術進行比較。

User privacy is an important issue in the cloud machine learning service. In this pa- per, we raise a new threat about the online machine learning service, which comes from outside superior authority. The authority may ask the user and the cloud to disclose se- crets and the authority can monitor the user behavior. We propose a protection approach called learnable encryption with deniability (LED), which can convince the outsider of the fake data and can protect the user privacy. Our use of learnable image encryption leads to a decrease in the accuracy of model predictions. We use distributed multi-model pre- diction queries to improve prediction accuracy. We also compared our scheme with other learnable encryption techniques.

[1] R. Agrawal and R. Srikant. Privacy-preserving data mining. SIGMOD Rec., 29(2):439–450, May 2000.
[2] S. I. Ahamed and V. Ravi. Privacy-preserving chaotic extreme learning machine with fully homomorphic encryption, 2022.
[3] E. Bisong. Google Colaboratory, pages 59–64. Apress, Berkeley, CA, 2019.
[4] J. W. Bos, K. Lauter, J. Loftus, and M. Naehrig. Improved security for a ring- based fully homomorphic encryption scheme. In M. Stam, editor, Cryptography and Coding, pages 45–64, Berlin, Heidelberg, 2013. Springer Berlin Heidelberg.
[5] R. Canetti, C. Dwork, M. Naor, and R. Ostrovsky. Deniable encryption. In B. S. K. Jr., editor, Advances in Cryptology - CRYPTO ’97, 17th Annual International Cryptology Conference, Santa Barbara, California, USA, August 17-21, 1997, Proceedings, volume 1294 of Lecture Notes in Computer Science, pages 90–104. Springer, 1997.
[6] H.Chabanne,A.deWargny,J.Milgram,C.Morel,andE.Prouff.Privacy-preserving classification on deep neural network. Cryptology ePrint Archive, Report 2017/035, 2017. https://eprint.iacr.org/2017/035.
[7] G. Chen, Q. Chen, X. Zhu, and Y. Chen. Encrypted image feature extraction by
privacy-preserving mfs. In 2018 7th International Conference on Digital Home (ICDH), pages 42–45, Nov. 2018.
[8] Y.FreundandR.E.Schapire.Adecision-theoreticgeneralizationofon-linelearning and an application to boosting. Journal of computer and system sciences, 55(1):119– 139, 1997.
[9] P. Gasti, G. Ateniese, and M. Blanton. Deniable cloud storage: sharing files via public-key deniability. In Proceedings of the 9th annual ACM workshop on Privacy in the electronic society, pages 31–42, 2010.
[10] R. Gilad-Bachrach, N. Dowlin, K. Laine, K. Lauter, M. Naehrig, and J. Wernsing. Cryptonets: Applying neural networks to encrypted data with high throughput and accuracy. In International Conference on Machine Learning, pages 201–210, 2016.[11] O. Goldreich, S. Micali, and A. Wigderson. How to play any mental game. In Proceedings of the Nineteenth Annual ACM Symposium on Theory of Computing, STOC ’87, pages 218–229, New York, NY, USA, 1987. ACM.
[12] K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016.
[13] H. Ito, Y. Kinoshita, M. Aprilpyone, and H. Kiya. Image to perturbation: An image transformation network for generating visually protected images for privacy- preserving deep neural networks. IEEE Access, 9:64629–64638, 2021.
[14] H. Kiya. Compressible and learnable encryption for untrusted cloud environments. CoRR, abs/1811.10254, 2018.
[15] A. Krizhevsky, G. Hinton, et al. Learning multiple layers of features from tiny images. 2009.
[16] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, and L. D. Jackel. Backpropagation applied to handwritten zip code recognition. Neural Computation, 1(4):541–551, 1989.
[17] J. Lee, H. Kang, Y. Lee, W. Choi, J. Eom, M. Deryabin, E. Lee, J. Lee, D. Yoo, Y. Kim, and J. No. Privacy-preserving machine learning with fully homomorphic encryption for deep neural network. CoRR, abs/2106.07229, 2021.
[18] J. Liu, M. Juuti, Y. Lu, and N. Asokan. Oblivious neural network predictions via minionn transformations. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, CCS ’17, pages 619–631, New York, NY, USA, 2017. ACM.
[19] K. Madono, M. Tanaka, M. Onishi, and T. Ogawa. Block-wise scrambled image recognition using adaptation network. CoRR, abs/2001.07761, 2020.
[20] I. Masi, Y. Wu, T. Hassner, and P. Natarajan. Deep face recognition: A survey. In 2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), pages 471–478, 2018.
[21] P. Mohassel and Y. Zhang. Secureml: A system for scalable privacy-preserving machine learning. In 2017 IEEE Symposium on Security and Privacy (SP), pages 19–38, May 2017.
[22] A. O’Neill, C. Peikert, and B. Waters. Bi-deniable public-key encryption. In P. Ro- gaway, editor, Advances in Cryptology – CRYPTO 2011, pages 525–542, Berlin, Heidelberg, 2011. Springer Berlin Heidelberg.
[23] R. Podschwadt, D. Takabi, and P. Hu. Sok: Privacy-preserving deep learning with homomorphic encryption. CoRR, abs/2112.12855, 2021.
[24] B. D. Rouhani, M. S. Riazi, and F. Koushanfar. Deepsecure: Scalable provably- secure deep learning. In Proceedings of the 55th Annual Design Automation Conference, DAC ’18, pages 2:1–2:6, New York, NY, USA, 2018. ACM.
[25] C. Shorten and T. M. Khoshgoftaar. A survey on image data augmentation for deep learning. Journal of big data, 6(1):1–48, 2019.
[26] W. Sirichotedumrong, Y. Kinoshita, and H. Kiya. Pixel-based image encryp- tion without key management for privacy-preserving deep neural networks. IEEE Access, 7:177844–177855, 2019.
[27] W. Sirichotedumrong and H. Kiya. A gan-based image transformation scheme for privacy-preserving deep neural networks. In 2020 28th European Signal Processing Conference (EUSIPCO), pages 745–749, 2021.
[28] M. Tanaka. Learnable image encryption. CoRR, abs/1804.00490, 2018.
[29] A. C. Yao. How to generate and exchange secrets. In 27th Annual Symposium on Foundations of Computer Science (sfcs 1986), pages 162–167, Oct. 1986.