HackerPapers

arXiv:1610.06918cs.CR

Learning to Protect Communications with Adversarial Neural Cryptography

Martín Abadi, David G. Andersen

Abstract

We ask whether neural networks can learn to use secret keys to protect information from other neural networks. Specifically, we focus on ensuring confidentiality properties in a multiagent system, and we specify those properties in terms of an adversary. Thus, a system may consist of neural networks named Alice and Bob, and we aim to limit what a third neural network named Eve learns from eavesdropping on the communication between Alice and Bob. We do not prescribe specific cryptographic algorithms to these neural networks; instead, we train end-to-end, adversarially. We demonstrate that the neural networks can learn how to perform forms of encryption and decryption, and also how to apply these operations selectively in order to meet confidentiality goals.

arXiv comments
15 pages
Journal reference
DOI
Subjects
cs.CRCryptography and Security cs.LGMachine Learning

On Hacker News

Linked by 5 identifiable commenters in 5 comments across 2 threads, from Oct 2016 to Aug 2017. Found in 5 HN submissions from Oct 2016 to Nov 2016. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments2 threads

  1. TFHE: Fast Fully-Homomorphic Encryption Over the Torus 2 comments

    • Do you think neural encryption is closer to encryption? GAN-style: One network encrypts while preserving structure, another network tries to reverse engineer to the original features. Edit: No specific sources for what Numerai is using, but in general…

      proofofstakeView on HNParent context

    • I've never heard of neural encryption before, do you have a good source for me? From the Wikipedia page for "Neural cryptography", it seems like there's some success in using NN's for cryptanalysis, but not for constructions... Edit: Do you mean the Google GAN experiment?…

      lkowalczView on HNParent context

  2. Google AI invents its own cryptographic algorithm 3 comments

    • This is the link to the research paper ("Learning to protect communications with adversarial neural cryptography"): https://arxiv.org/abs/1610.06918

      trungaczneView on HNParent context

    • Please don't editorialize. Finding the best parameters that optimize an objective function can hardly be called "re-writing its own code." Granted, the article is just as guilty. (Edit: The article also doesn't mention self-modifying code so I'm not sure where you're getting…

      gcrView on HNParent context

    • The paper [1] is by Abadi [2], which is a pretty big name in crypto. So I can assume it's something with value, wether the press can or can't convey that is another story. [1] https://arxiv.org/pdf/1610.06918v1.pdf [2] https://en.wikipedia.org/wiki/Mart%C3%ADn_Abadi

      ecesenaView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.