“I don't really believe the security model is realistic (how do we know NN's are really that effective as adversaries?), but at least there is a model”
Learning to Protect Communications with Adversarial Neural Cryptography
让神经网络 Alice 与 Bob 通过对抗训练学会对抗窃听者 Eve 的加密
论文研究神经网络能否利用密钥保护信息不被另一个网络窃听:Alice 和 Bob 通信,Eve 窃听,三者通过端到端的对抗训练学习,不预设任何具体密码算法。结果表明网络可以学会加密和解密,并能有选择地使用这些操作来满足保密目标。
HN 上的人把它当作神经网络能自学加密的新闻背后的原始论文来引用,并以作者 Abadi 是知名密码学者为其背书,同时也有人质疑其安全模型不够现实。
学习路径
读这篇之前,按顺序读书架上的 2 篇 · 参考文献来自 OpenAlex,顺序和要点由 Claude Code 排出先看 GAN 的对抗训练技巧,再读两个网络与一个窃听网络之间的对抗训练如何学出加密。
- 1
- 2
神经网络与加密数据结合的先例,同属神经网络与密码学交叉
读网络在加密数据上做计算的设定,对比让网络自己学加密
- 3Learning to Protect Communications with Adversarial Neural Cryptography
目标论文:让 Alice 与 Bob 通过对抗训练学会对抗窃听者 Eve 的加密
读三方网络的损失设计,以及学到的加密方式如何评估
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“This is the link to the research paper ("Learning to protect communications with adversarial neural cryptography"):”
“For the curious, here is the preprint: Learning to Protect Communications with Adversarial Neural Cryptography”
“The paper [1] is by Abadi [2], which is a pretty big name in crypto.”
“No specific sources for what Numerai is using, but in general: Learning to Protect Communications with Adversarial Neural Cryptography”
- Learning to Protect Communications with Adversarial Neural Cryptography ↗dooglius · 2016-10-27 · 3 分
- Google's Neural Network Invents Its Own Encryption ↗sdebrule · 2016-10-27 · 3 分
- Google Brain creates its own encryption methodology [paper] ↗sarreph · 2016-10-28 · 3 分
- Learning to Protect Communications with Adversarial Neural Cryptography ↗rabbitcore · 2016-11-08 · 2 分
- Learning to Protect Communications with Adversarial Neural Cryptography ↗thibaut_barrere · 2016-11-13 · 4 分