{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/crypto-nets-neural-networks-over-encrypted","title":"Crypto-Nets: Neural Networks over Encrypted Data","arxiv_id":"1412.6181","date":"2014-12-18","proceeding":null,"authors":["Pengtao Xie","Misha Bilenko","Tom Finley","Ran Gilad-Bachrach","Kristin Lauter","Michael Naehrig"],"abstract":"The problem we address is the following: how can a user employ a predictive\nmodel that is held by a third party, without compromising private information.\nFor example, a hospital may wish to use a cloud service to predict the\nreadmission risk of a patient. However, due to regulations, the patient's\nmedical files cannot be revealed. The goal is to make an inference using the\nmodel, without jeopardizing the accuracy of the prediction or the privacy of\nthe data.\n  To achieve high accuracy, we use neural networks, which have been shown to\noutperform other learning models for many tasks. To achieve the privacy\nrequirements, we use homomorphic encryption in the following protocol: the data\nowner encrypts the data and sends the ciphertexts to the third party to obtain\na prediction from a trained model. The model operates on these ciphertexts and\nsends back the encrypted prediction. In this protocol, not only the data\nremains private, even the values predicted are available only to the data\nowner.\n  Using homomorphic encryption and modifications to the activation functions\nand training algorithms of neural networks, we show that it is protocol is\npossible and may be feasible. This method paves the way to build a secure\ncloud-based neural network prediction services without invading users' privacy.","url_abs":"http://arxiv.org/abs/1412.6181v2","url_pdf":"http://arxiv.org/pdf/1412.6181v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"crypto-nets-neural-networks-over-encrypted","repo_url":"https://github.com/microsoft/CryptoNets","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"prediction","task_name":"Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1412.6181","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}