Efficient Skip Connections Realization for Secure Inference on Encrypted Data
Pages 65 - 73
Abstract
Homomorphic Encryption (HE) is a cryptographic tool that allows performing computation under encryption, which is used by many privacy-preserving machine learning solutions, for example, to perform secure classification. Modern deep learning applications yield good performance for example in image processing tasks benchmarks by including many skip connections. The latter appears to be very costly when attempting to execute model inference under HE. In this paper, we show that by replacing (mid-term) skip connections with (short-term) Dirac parameterization and (long-term) shared-source skip connection we were able to reduce the skip connections burden for HE-based solutions, achieving computing power improvement for the sameaccuracy.
References
[1]
Aharoni, E., et al.: HeLayers: A Tile Tensors Framework for Large Neural Networks on Encrypted Data. CoRR abs/2011.0 (2020). arXiv:2011.01805
[2]
Albrecht, M., et al.: Homomorphic encryption security standard. Technical report, HomomorphicEncryption.org, Toronto, Canada (2018). https://HomomorphicEncryption.org
[3]
Balduzzi, D., Frean, M., Leary, L., Lewis, J.P., Ma, K.W.D., McWilliams, B.: The shattered gradients problem: If resnets are the answer, then what is the question? In: Precup, D., Teh, Y.W. (eds.) Proceedings of the 34th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 70, pp. 342–350. PMLR (2017). https://proceedings.mlr.press/v70/balduzzi17b.html
[4]
Baruch, M., et al.: Sensitive Tuning of Large Scale CNNs for E2E Secure Prediction using HE (2023)
[5]
Centers for Medicare & Medicaid Services: The Health Insurance Portability and Accountability Act of 1996 (HIPAA) (1996). https://www.hhs.gov/hipaa/
[6]
Cheon, J.H., Han, K., Kim, A., Kim, M., Song, Y.: A full RNS variant of approximate homomorphic encryption. In: Cid, C., Jacobson Jr., M.J. (eds.) Selected Areas in Cryptography - SAC 2018, SAC 2018. Lecture Notes in Computer Science, vol. 11349, pp. 347–368. Springer, Cham (2019).
[7]
Cheon JH, Kim A, Kim M, and Song Y Takagi T and Peyrin T Homomorphic encryption for arithmetic of approximate numbers Advances in Cryptology – ASIACRYPT 2017 2017 Cham Springer 409-437
[8]
CryptoLab: HEaaN: Homomorphic Encryption for Arithmetic of Approximate Numbers (2022). https://www.cryptolab.co.kr/eng/product/heaan.php
[9]
Dosovitskiy, A., et al.: An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020)
[10]
EU General Data Protection Regulation: Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). Official Journal of the European Union 119 (2016). http://data.europa.eu/eli/reg/2016/679/oj
[11]
Gartner: Gartner identifies top security and risk management trends for 2021. Technical report (2021). https://www.gartner.com/en/newsroom/press-releases/2021-03-23-gartner-identifies-top-security-and-risk-management-t
[12]
Gilad Bachrach, R., Dowlin, N., Laine, K., Lauter, K., Naehrig, M., Wernsing, J.: Cryptonets: applying neural networks to encrypted data with high throughput and accuracy. In: International Conference on Machine Learning, pp. 201–210 (2016). http://proceedings.mlr.press/v48/gilad-bachrach16.pdf
[13]
Halevi, S.: Homomorphic Encryption. In: Lindell, Y. (ed.) Tutorials on the Foundations of Cryptography: Dedicated to Oded Goldreich, pp. 219–276. Springer, Cham (2017).
[14]
He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016). https://openaccess.thecvf.com/content_cvpr_2016/html/He_Deep_Residual_Learning_CVPR_2016_paper.html
[15]
IBM: HELayers SDK with a Python API for x86 (2021). https://hub.docker.com/r/ibmcom/helayers-pylab
[16]
Lee, E., et al.: Low-complexity deep convolutional neural networks on fully homomorphic encryption using multiplexed parallel convolutions. In: Chaudhuri, K., Jegelka, S., Song, L., Szepesvari, C., Niu, G., Sabato, S. (eds.) Proceedings of the 39th International Conference on Machine Learning, vol. 162, pp. 12403–12422. PMLR (2022). https://proceedings.mlr.press/v162/lee22e.html
[17]
Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.: A convnet for the 2020s. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11976–11986 (2022)
[18]
van den Oord, A., et al.: Wavenet: a generative model for raw audio. In: 9th ISCA Speech Synthesis Workshop, pp. 125–125 (2016). https://www.isca-speech.org/archive_v0/SSW_2016/abstracts/ssw9_DS-4_van_den_Oord.html
[19]
Oyedotun, O.K., Shabayek, A.E.R., Aouada, D., Ottersten, B.: Going deeper with neural networks without skip connections. In: 2020 IEEE International Conference on Image Processing (ICIP), pp. 1756–1760 (2020).
[20]
Pascanu, R., Mikolov, T., Bengio, Y.: On the difficulty of training recurrent neural networks. In: Dasgupta, S., McAllester, D. (eds.) Proceedings of the 30th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 28, pp. 1310–1318. PMLR (2013). https://proceedings.mlr.press/v28/pascanu13.html
[21]
Radford, A., Narasimhan, K., Salimans, T., Sutskever, I., et al.: Improving language understanding by generative pre-training (2018). https://www.cs.ubc.ca/amuham01/LING530/papers/radford2018improving.pdf
[22]
Tai, Y., Yang, J., Liu, X.: Image super-resolution via deep recursive residual network. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2790–2798 (2017).
[23]
The HEBench Organization: HEBench (2022). https://hebench.github.io/
[24]
Vaswani, A., et al.: Attention is all you need. In: Advances in Neural Information Processing Systems, vol. 30 (2017). https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf
[25]
Zagoruyko, S., Komodakis, N.: Diracnets: Training very deep neural networks without skip-connections (2017). arXiv:1706.00388
Recommendations
Secure Data Devolution: Practical Re-encryption with Auxiliary Data in LWE-based Somewhat Homomorphic Encryption
SCC '15: Proceedings of the 3rd International Workshop on Security in Cloud ComputingHomomorphic encryption can support meaningful operations on encrypted data, and hence it enables users to outsource their data in encrypted format to cloud services. However, homomorphic encryption cannot operate on ciphertexts with different keys in ...
An efficient secure proxy verifiably encrypted signature scheme
Verifiably encrypted signature is an important cryptographic primitive, it can convince a verifier that a given ciphertext is an encryption of signature on a given message. It is often used as a building block to construct an optimistic fair exchange. ...
Chosen ciphertext secure keyed-homomorphic public-key cryptosystems
In homomorphic encryption schemes, anyone can perform homomorphic operations, and therefore, it is difficult to manage when, where and by whom they are performed. In addition, the property that anyone can "freely" perform the operation inevitably means ...
Comments
Please enable JavaScript to view thecomments powered by Disqus.Information & Contributors
Information
Published In
Jun 2023
538 pages
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.
Publisher
Springer-Verlag
Berlin, Heidelberg
Publication History
Published: 29 June 2023
Author Tags
Qualifiers
- Article
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 0Total Downloads
- Downloads (Last 12 months)0
- Downloads (Last 6 weeks)0
Reflects downloads up to 14 Feb 2025