Abstract
The foreseen growing role of outsourced machine learning services is raising concerns about the privacy of user data. This paper proposes a variety of protocols for privacy-preserving regression and classification that (i) only require additively homomorphic encryption algorithms, (ii) limit interactions to a mere request and response, and (iii) that can be used directly for important machine-learning algorithms such as logistic regression and SVM classification. The basic protocols are then extended and applied to simple feed-forward neural networks.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Given , the server obtains as if \(\eta _i = 0\), and as if \(\eta _i = 1\).
References
Abu-Mostafa, Y.S., Magdon-Ismail, M., Lin, H.T.: Learning From Data: A Short Course. AMLbook.com, New York (2012). http://amlbook.com
Agrawal, R., Srikant, R.: Privacy-preserving data mining. ACM SIGMOD Rec. 29(2), 439–450 (2000). https://doi.org/10.1145/335191.335438
Barni, M., Orlandi, C., Piva, A.: A privacy-preserving protocol for neural-network-based computation. In: MM&Sec 2006, pp. 146–151. ACM (2006). https://doi.org/10.1145/1161366.1161393
Bos, J.W., Lauter, K., Naehrig, M.: Private predictive analysis on encrypted medical data. J. Biomed. Inf. 50, 234–243 (2014). https://doi.org/10.1016/j.jbi.2014.04.003
Bost, R., Popa, R.A., Tu, S., Goldwasser, S.: Machine learning classification over encrypted data. In: NDSS 2015. The Internet Society (2015). https://doi.org/10.14722/ndss.2015.23241
Damgård, I., Geisler, M., Krøigaard, M.: Homomorphic encryption and secure comparison. Int. J. Appl. Cryptogr. 1(1), 22–31 (2008). https://doi.org/10.1504/IJACT.2008.017048
Damgård, I., Geisler, M., Krøigaard, M.: A correction to ‘efficient and secure comparison for on-line auctions’. Int. J. Appl. Cryptogr. 1(4), 323–324 (2009). https://doi.org/10.1504/IJACT.2009.028031
Dwork, C., Feldman, V.: Privacy-preserving prediction. In: COLT 2018. PMLR, vol. 75, pp. 1693–1702. PMLR (2018). http://proceedings.mlr.press/v75/dwork18a/dwork18a.pdf
Erkin, Z., Franz, M., Guajardo, J., Katzenbeisser, S., Lagendijk, I., Toft, T.: Privacy-preserving face recognition. In: Goldberg, I., Atallah, M.J. (eds.) PETS 2009. LNCS, vol. 5672, pp. 235–253. Springer, Heidelberg (2009). https://doi.org/10.1007/978-3-642-03168-7_14
Glorot, X., Bordes, A., Bengjio, Y.: Deep sparse rectifier neural networks. In: AISTAT 2011. PMLR, vol. 15, pp. 315–323. PMLR (2011). http://proceedings.mlr.press/v15/glorot11a/glorot11a.pdf
Goethals, B., Laur, S., Lipmaa, H., Mielikäinen, T.: On private scalar product computation for privacy-preserving data mining. In: Park, C., Chee, S. (eds.) ICISC 2004. LNCS, vol. 3506, pp. 104–120. Springer, Heidelberg (2005). https://doi.org/10.1007/11496618_9
Goldwasser, S., Micali, S.: Probabilistic encryption. J. Comput. Syst. Sci. 28(2), 270–299 (1984). https://doi.org/10.1016/0022-0000(84)90070-9
Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning. Springer Series in Statistics, 2nd edn. Springer, New York (2009). https://doi.org/10.1007/978-0-387-84858-7
Hubara, I., Courbariaux, M., Soudry, D., El-Yaniv, R., Bengio, Y.: Binarized neural networks. In: NISP 2016, pp. 4107–4115. Curran Associates, Inc. http://papers.nips.cc/paper/6573-binarized-neural-networks.pdf
Joye, M., Salehi, F.: Private yet efficient decision tree evaluation. In: Kerschbaum, F., Paraboschi, S. (eds.) DBSec 2018. LNCS, vol. 10980, pp. 243–259. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-95729-6_16
Kim, M., Song, Y., Wang, S., Xia, Y., Jiang, X.: Secure logistic regression based on homomorphic encryption: design and evaluation. JMIR Med. Inform. 6(2), e19 (2018). https://doi.org/10.2196/medinform.8805
Lindell, Y., Pinkas, B.: Privacy preserving data mining. In: Bellare, M. (ed.) CRYPTO 2000. LNCS, vol. 1880, pp. 36–54. Springer, Heidelberg (2000). https://doi.org/10.1007/3-540-44598-6_3
Mohassel, P., Zhang, Y.: SecureML: A system for scalable privacy-preserving machine learning. In: IEEE S&P 2017, pp. 19–38. IEEE Computer Society (2017). https://doi.org/10.1109/SP.2017.12
Tramèr, F., Zhang, F., Juels, A., Reiter, M.K., Ristenpart, T.: Stealing machine learning models via prediction APIs. In: USENIX Security 2016, pp. 601–618. USENIX Association (2016). https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_tramer.pdf
Veugen, T.: Improving the DGK comparison protocol. In: WIFS 2012, pp. 49–54. IEEE (2012). https://doi.org/10.1109/WIFS.2012.6412624
Zhang, J., Wang, X., Yiu, S.M., Jiang, Z.L., Li, J.: Secure dot product of outsourced encrypted vectors and its application to SVM. In: SCC@AsiaCCS 2017, pp. 75–82. ACM (2017). https://doi.org/10.1145/3055259.3055270
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
A A More Private Protocols
A A More Private Protocols
1.1 A.1 Linear Regression
As seen in Sect. 2.1, linear regression produces estimates using the identity map for g: . Since is linear, given an encryption of \({{\varvec{x}}}\), the value of can be homomorphically evaluated, in a provably secure way [11].
Therefore, the client encrypts its feature vector \({{\varvec{x}}}\) under its public key and sends to the server. Using \({\varvec{\theta }}\), the server then computes and returns it the client. Finally, the client decrypts and gets the output . This is straightforward and only requires one round of communication.
1.2 A.2 SVM Classification
A Naïve Protocol. A client holding a private feature vector \({{\varvec{x}}}\) wishes to evaluate where \({\varvec{\theta }}\) parametrises an SVM classification model. In the primal approach, the client can encrypt \({{\varvec{x}}}\) and send to the server. Next, the server computes for some random mask \(\mu \) and sends to the client. The client decrypts and recovers \(\eta \). Finally, the client and the server engage in a private comparison protocol with respective inputs \(\eta \) and \(\mu \), and the client deduces the sign of from the resulting comparison bit \([\mu \leqslant \eta ]\).
There are two issues. If we use the DGK+ protocol for the private comparison, at least one extra exchange from the server to the client is needed for the client to get \([\mu \leqslant \eta ]\). This can be fixed by considering the dual approach. A second, more problematic, issue is that the decryption of yields \(\eta \) as an element of \(\mathcal {M}\cong \mathbb {Z}/M\mathbb {Z}\), which is not necessarily equivalent to the integer . Note that if the inner product can take any value in \(\mathcal {M}\), selecting a smaller value for \(\mu \in \mathcal {M}\) to prevent the modular reduction does not solve the issue because the value of \(\eta \) may then leak information on .
A Heuristic Protocol (Dual Approach). The bandwidth usage with the heuristic comparison protocol (cf. Fig. 5) could be even reduced to one ciphertext and a single bit with the dual approach. From the published encrypted model , the client could homomorphically compute and send to the server for random \(\lambda , \mu \in \mathcal {B}\) with . The server would then decrypt \(\mathfrak {t}^*\), obtain \(t^*\), compute \({\smash {\delta }_{\scriptscriptstyle S}}= \frac{1}{2}(1-\mathrm{sign}(t^*))\), and return \({\smash {\delta }_{\scriptscriptstyle S}}\) to the client. Analogously to the primal approach, the output class is obtained by the client as \(\hat{y}= (-1)^{\smash {\delta }_{\scriptscriptstyle S}}\cdot \mathrm{sign}(\lambda )\). However, and contrarily to the primal approach, the potential information leakage resulting from \(t^*\)—in this case on \({{\varvec{x}}}\)—is now on the server’s side, which is in contradiction with our Requirement 1 (input confidentiality). We do not further discuss this variant.
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Joye, M., Petitcolas, F. (2019). PINFER: Privacy-Preserving Inference. In: Pérez-Solà, C., Navarro-Arribas, G., Biryukov, A., Garcia-Alfaro, J. (eds) Data Privacy Management, Cryptocurrencies and Blockchain Technology. DPM CBT 2019 2019. Lecture Notes in Computer Science(), vol 11737. Springer, Cham. https://doi.org/10.1007/978-3-030-31500-9_1
Download citation
DOI: https://doi.org/10.1007/978-3-030-31500-9_1
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-31499-6
Online ISBN: 978-3-030-31500-9
eBook Packages: Computer ScienceComputer Science (R0)