Privacy preserving extreme learning machine using additively homomorphic encryption

Recently, computational outsourcing using cloud services is getting popular for big data analysis, and many cloud sourcing providers provide machine learning platforms where we can perform various prediction and classification tasks very easily. On the other hand, there still remains a big hurdle to analyze personal big data on cloud services because the leakage of personal information is a critical issue. As a remedy for this, we propose a privacy preserving machine learning algorithm for Extreme Learning Machine (PP-ELM), which can learn from data encrypted with an additively homomorphic encryption. In the proposed outsourcing method, we consider a three-participants model consisting of data contributors, outsourced server, and data analyst. A data contributor preprocesses and encrypts data, and an outsourced server receives encrypted data and calculate hidden layer outputs using additive operations. Then, a data analyst receives the hidden outputs of ELM from the outsourced server and they are used to obtain ELM connection weights. Since the proposed outsourcing model can learn ELM over encrypted data, it is expected to mitigate a hurdle to deal with personal data on cloud services. In addition, the proposed PP-ELM allows us to learn multiple sources of personal data in a secure way, and this might lead to a better solution for a practical problem than before.

[1]  T. Elgamal.  A public key cryptosystem and a signature scheme based on discrete logarithms , 1984, CRYPTO 1984.

[2]  Hongming Zhou,et al.  Extreme Learning Machine for Regression and Multiclass Classification , 2012, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics).

[3]  Michael Naehrig,et al.  CryptoNets: applying neural networks to encrypted data with high throughput and accuracy , 2016, ICML 2016.

[4]  Craig Gentry,et al.  Fully homomorphic encryption using ideal lattices , 2009, STOC '09.

[5]  Shucheng Yu,et al.  Privacy Preserving Back-Propagation Neural Network Learning Made Practical with Cloud Computing , 2014, IEEE Transactions on Parallel and Distributed Systems.

[6]  Cynthia Dwork,et al.  Calibrating Noise to Sensitivity in Private Data Analysis , 2006, TCC.

[7]  Shiho Moriai,et al.  Privacy-Preserving Deep Learning: Revisited and Enhanced , 2017, ATIS.

[8]  Kamalika Chaudhuri,et al.  Privacy-preserving logistic regression , 2008, NIPS.

[9]  Pascal Paillier,et al.  Public-Key Cryptosystems Based on Composite Degree Residuosity Classes , 1999, EUROCRYPT.

[10]  F ROSENBLATT,et al.  The perceptron: a probabilistic model for information storage and organization in the brain. , 1958, Psychological review.

[11]  Taher ElGamal,et al.  A public key cyryptosystem and signature scheme based on discrete logarithms , 1985 .

[12]  Yoshinori Aono,et al.  Fast and Secure Linear Regression and Biometric Authentication with Security Update , 2015, IACR Cryptol. ePrint Arch..

[13]  Ali Miri,et al.  Privacy-preserving back-propagation and extreme learning machine algorithms , 2012, Data Knowl. Eng..

[14]  Adi Shamir,et al.  A method for obtaining digital signatures and public-key cryptosystems , 1978, CACM.

[15]  Adi Shamir,et al.  A method for obtaining digital signatures and public-key cryptosystems , 1978, CACM.

[16]  Yoshinori Aono,et al.  Scalable and Secure Logistic Regression via Homomorphic Encryption , 2016, IACR Cryptol. ePrint Arch..

[17]  Shafi Goldwasser,et al.  Machine Learning Classification over Encrypted Data , 2015, NDSS.

[18]  Ronald L. Rivest,et al.  ON DATA BANKS AND PRIVACY HOMOMORPHISMS , 1978 .

[19]  Ehsan Hesamifard,et al.  Preserving Multi-party Machine Learning with Homomorphic Encryption , 2016 .