Neural network learning is a technique that is used to solve problems of classification, prediction, clustering, modelling based on variety of data inputs in the form of structured, semi-structured and unstructured data. Learning accuracy is considered as key performance index in these neural network based learning algorithms. Many organization that involves huge amount of data would want to outsource it to cloud for artificial intelligence based services. Various organization who wish to train neural network model on their complex and huge data usually outsource the learning model on cloud. Outsourcing of learning model on cloud creates security concerns for input data and the learned model. In this paper, we propose a practical system that will train a neural network model that is encrypted during training process. The training is performed on the unencrypted data. The output of the system is a neural network model that possesses two properties. First, neural network model is protected from the malicious users, hence allows the users to train the model in insecure environments at no cost of risk. Second, the neural network model can make only encrypted predictions. We make use of homomorphic encryption techniques to fulfill the objectives and test our results on sentiment analysis dataset.

Homomorphic encryption, neural network

[1] S. Chow, Y. He, and et al. Spice - simple privacy-preserving identity-management for cloud environment. In ACNS 2012, volume 7341 of Lecture Notes in Computer Science. Springer, 2012.
[2] Privacy Preserving Back-Propagation Neural Network Learning Made Practical with Cloud Computing. IEEE Transactions on Parallel and Distributed Systems, Vol. 25, No. 1, January 2014.
[3] N. Schlitter, A Protocol for Privacy Preserving Neural Network Learning on Horizontal Partitioned Data, Proc. Privacy Statistics in Databases (PSD 08), Sept. 2008
[4] Erich Schikuta and Erwin Mann, N2Sky - Neural Networks as Services in the Clouds. arXiv:1401.2468v1 [cs.NE] 10 Jan 2014.
[5] T. Chen and S. Zhong,Privacy-Preserving Backpropagation Neural Network Learning, IEEE Trans. Neural Network, vol. 20, no. 10, pp. 1554-1564, Oct. 2009.
[6] Mohammad Ali Kadampur, Somayajulu D.V.L.N. A Noise Addition Scheme in Decision Tree for Privacy Preserving Data Mining, Journal of Computing, Volumen 2, Issue 1, January 2010, ISSN 2151-9617
[7] Yong Liu, Yeming Xiao, Li Wang, Jielin Pan, Yonghong Yan. Parallel Implementation of Neural Networks Training on Graphic Processing Unit, 2012 5th International Conference on BioMedical Engineering and Informatics (BMEI 2012)
[8] Pelin Angin, Bharat Bhargava, Rohit Ranchal, Noopur Singh. An Entity-centric Approach for Privacy and Identity Management in Cloud Computing, 2010 29th IEEE International Symposium on Reliable Distributed Systems.
[9] Scretan J, Georgiopoulos, M. A privacy preserving probabilistic neural network for horizontally partitioned databases. International Joint Conference on Neural Networks. Aug 2007.
[10] Barni M, Failla P, Sadeghi A. Privacy Preserving ECG Classification with branching programs and neural networks.IEEE Transaction. Information Forensics and Security. Volume 6, Issue 2, June 2011.
[11] Samet S. Privacy Preserving protocols for perceptron learning algorithm in neural networks. IEEE Conference on Intelligent Systems, Sept 2008.
[12] Mahmoud Barhamgi, Arosha K. Bandara, and Yijun Yu, Protecting Privacy in the Cloud: Current Practices, Future Directions, Computer IEEE Society February 2016.
[13] Majid Bashir Malik, A model for Privacy Preserving in Data Mining using Soft Computing Techniques. March 2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom).
[14] Reza Shokri, Privacy-Preserving Deep Learning,, 2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) Oct 2015.
[15] Nathan Dowlin, Ran Gilad-Bachrach, Kim Laine, Kristin Lauter, Michael Naehrig and John Wernsing, CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy 29 December 2015
[16] Ryan Hayward , Chia-Chu Chiang, Parallelizing fully homomorphic encryption for a cloud environment. Journal of Applied Research and Technology 13 (2015) 245-252
[17] Bengio. Learning deep architectures for AI. Foundations and trends in machine learning, 2(1):1– 127, 2009.
[18] L. Deng. A tutorial survey of architectures, algorithms, and applications for deep learning. APSIPA Trans. Signal and Information Processing, 3, 2014.
[19] A. Graves, A.-R. Mohamed, and G. Hinton. Speech recognition with deep recurrent neural networks. In ICASSP , 2013.
[20] Hannun, C. Case, J. Casper, B. Catanzaro, G. Diamos, E. Elsen, R. Prenger, S. Satheesh, S. Sengupta, A. Coates, et al. Deepspeech: Scaling up end-to-end speech recognition. arXiv:1412.5567 , 2014.
[21] G. Hinton, L. Deng, D. Yu, G. Dahl, A.-r. Mohamed, N. Jaitly, A. Senior, V. Vanhoucke, P. Nguyen, T. Sainath, et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. Signal Processing Magazine , 29(6):82–97, 2012.
[22] A. Krizhevsky, I. Sutskever, and G. Hinton. Imagenet classification with deep convolutional neural networks. In NIPS , 2012.
[23] P. Simard, D. Steinkraus, and J. Platt. Best practices for convolutional neural networks applied to visual document analysis. In Document Analysis and Recognition , 2013.
[24] Y. Taigman, M. Yang, M. Ranzato, and L. Wolf. Deepface: Closing the gap to human-level performance in face verification. In CVPR , 2014.
[25] Angel Yu, Wai Lok Lai, James Pay or Efficient Integer Vector Homomorphic Encryption, May 2015.