Abstract
Problems involving the classification of sensitive cloud data, such as medical or financial records, can be solved using a convolutional neural network (CNN) approach. Data privacy and security, however, require extra care. Homomorphic encryption is a potential security solution to allow CNNs to classify cypher data. The convolution layer, for example, is a computation that can be fully homomorphically evaluated; however, non-linear functions, such as the activation layer, cannot be fully homomorphically assessed since they do not follow a linear pattern. This paper uses low-degree polynomials to approximate these non-linear functions. We compare various polynomial approximation approaches to achieve the optimal approximation function for activation functions such as Relu. Finally, we use a proposed CNN and the MNIST dataset to demonstrate the effectiveness of our approximation functions. The resulting proposed CNN achieved a 99.80% accuracy rate.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Chou, E., Beal, J., Levy, D., Yeung, S., Haque, A., Fei-Fei, L.: Faster CryptoNets: leveraging sparsity for real-world encrypted inference. arXiv, November 2018. https://arxiv.org/abs/1811.09953. Accessed 14 Nov 2021
Denning, D.E.R.: Cryptography and data security (1982)
Yassein, H.R., Al-Saidi, N.M.G., Farhan, A.K.: A new NTRU cryptosystem outperforms three highly secured NTRU-analog systems through an innovational algebraic structure. J. Discret. Math. Sci. Cryptogr. (2020). https://doi.org/10.1080/09720529.2020.1741218
Kadhim, A.F., Kamal, Z.A.: Generating dynamic S-box based on particle swarm optimization and chaos theory for AES. Iraqi J. Sci. 59(3), 1733–1745 (2018). https://doi.org/10.24996/IJS.2018.59.3C.18
Alhudhaif, A., Ahmad, M., Alkhayyat, A., Tsafack, N., Farhan, A.K., Ahmed, R.: Block cipher nonlinear confusion components based on new 5-D hyperchaotic system. IEEE Access 9, 87686–87696 (2021). https://doi.org/10.1109/ACCESS.2021.3090163
Zahid, A.H., et al.: Efficient dynamic S-box generation using linear trigonometric transformation for security applications. IEEE Access 9, 98460–98475 (2021). https://doi.org/10.1109/ACCESS.2021.3095618
Podschwadt, R., Takagi, D., Hu, P.: SoK: privacy-preserving deep learning with homomorphic encryption (2021). http://arxiv.org/abs/2112.12855
Yao, A.C.C.: How to generate and exchange secrets. In: Annual Symposium Foundations of Computer Science, no. 1, pp. 162–167 (1986). https://doi.org/10.1109/sfcs.1986.25
Shamir, A.: How to share a secret. Commun. ACM 22(11), 612–613 (1979). https://doi.org/10.1145/359168.359176
Gentry, C.S.: Fully homomorphic encryption from approximate ideal lattices. Ruan Jian Xue Bao J. Softw. 26(10), 2696–2719 (2015). https://doi.org/10.13328/j.cnki.Jos.004808
Cheon, J.H., Han, K., Kim, A., Kim, M., Song, Y.: Bootstrapping for approximate homomorphic encryption. In: Nielsen, J.B., Rijmen, V. (eds.) EUROCRYPT 2018. LNCS, vol. 10820, pp. 360–384. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-78381-9_14
McKeen, F., et al.: Innovative instructions and software model for isolated execution. In: Proceeding of the 2nd International Workshop HASP (2013). https://doi.org/10.1145/2487726.2488368
Obla, S., Gong, X., Aloufi, A., Hu, P., Takabi, D.: Effective activation functions for homomorphic evaluation of deep neural networks. IEEE Access 8, 153098–153112 (2020). https://doi.org/10.1109/ACCESS.2020.3017436
Dowlin, N., et al.: CryptoNets: applying neural networks to encrypted data with high throughput and accuracy - Microsoft research. Microsoft Res. TechReport 48, 1–12 (2016). http://proceedings.mlr.press/v48/gilad-bachrach16.pdf. http://research.microsoft.com/apps/pubs/?id=260989
Chabanne, H.: Privacy-preserving classification on deep neural network. Med. Today 4(7), 59–64 (2017)
Hesamifard, E., Takagi, H., Ghasemi, M.: CryptoDL: deep neural networks over encrypted data, pp. 1–21 (2017). http://arxiv.org/abs/1711.05189
Bourse, F., Minelli, M., Minihold, M., Paillier, P.: Fast homomorphic evaluation of deep discretized neural networks. In: Shacham, H., Boldyreva, A. (eds.) CRYPTO 2018. LNCS, vol. 10993, pp. 483–512. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-96878-0_17
Deng, L.: The MNIST database of handwritten digit images for machine learning research. IEEE Signal Process. Mag. 29(6), 141–142 (2012). https://doi.org/10.1109/MSP.2012.2211477
Ohn, I., Kim, Y.: Smooth function approximation by deep neural networks with general activation functions. Entropy 21(7), 1–24 (2019). https://doi.org/10.3390/e21070627
Maennel, H., Bousquet, O., Gelly, S.: Gradient descent quantizes ReLU network features, arXiv, vol. abs/1803.0 (2018)
Harrou, F., Zeroual, A., Sun, Y.: Calcuas. In: Proceedings of the American Control Conference, vol. 2018-June, pp. 604–609 (2018). https://doi.org/10.23919/ACC.2018.8431387
Obla, S., Gong, X., Aloufi, A., Hu, P., Takabi, D.: Effective activation functions for homomorphic evaluation of deep neural networks (2020)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Alsaedi, E.M., Farhan, A.K., Falah, M.W., Oleiwi, B.K. (2022). Classification of Encrypted Data Using Deep Learning and Legendre Polynomials. In: Daimi, K., Al Sadoon, A. (eds) Proceedings of the ICR’22 International Conference on Innovations in Computing Research. ICR 2022. Advances in Intelligent Systems and Computing, vol 1431. Springer, Cham. https://doi.org/10.1007/978-3-031-14054-9_31
Download citation
DOI: https://doi.org/10.1007/978-3-031-14054-9_31
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-14053-2
Online ISBN: 978-3-031-14054-9
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)