安全与密码学2021apple.com
Apple PSI 系统：安全协议与分析
The Apple Psi System
专业 · 约 90 分钟读完前置知识：私有集合求交（PSI）、密码学安全证明、椭圆曲线密码学
导读
这是苹果发布的 PDF，详细说明其儿童安全（CSAM 检测）系统所采用的 PSI（私有集合求交）协议及其形式化安全分析。核心做法是：客户端在本地把照片的感知哈希与服务器下发的哈希数据库做私有集合求交，只有当匹配数量超过阈值时，才向服务器披露可用于人工审核的信息。文档重点论证该协议在密码学层面的隐私与正确性保证。
HN 关注点：的焦点是：密码学证明再严谨，也改变不了这是给所有 iPhone 装上一套不可审计的客户端扫描系统的事实，讨论集中在威胁模型、误报与权力滥用上。
后来怎么样
该文档公开后引发了对客户端 CSAM 扫描的广泛争议，苹果最终搁置了该检测计划，并使其成为隐私增强技术与大规模监控边界讨论的典型案例。
HN 评论精选
这几条评论里：质疑 4
质疑技术再严谨也改变不了本质：这是在用户设备上强制部署的间谍软件，不应存在。
这非常令人恼火。他们提供了所有这些信息，说明这个运行在我们设备上的程序非常安全——前提是我们不是坏人；在人类编写的算法触发人类设定的阈值之前，苹果和执法部门看不到我们照片的任何信息。技术上这应该站得住脚，因为多位密码学专家已经验证了证明。但这些都不重要。Apple PSI 系统就是间谍软件。他们提供所有这些信息，只是为了给在我们的设备上安装间谍软件找理由。他们试图在我们的设备上装间谍软件，看我们会不会因此被送进监狱。这才是重点，这就是最终效果。苹果在为把间谍软件装到他们所有手机上做辩护。任何关于这套针对你部署的间谍软件的技术优点的讨论都偏离了重点。它根本不应该存在。
drenvuk查看英文原文
This is incredibly annoying. They're providing all of this information which says that this program that runs on our devices is incredibly safe if we're not bad people. That the people at Apple and then law enforcement don't see anything about our photos until there is a human set threshold hit by a human programmed algorithm. Technically, this should be sound since the proof has checked out with multiple cryptography experts. But none of that matters. The Apple PSI System is spyware. They are providing all of this info to justify putting spyware on our devices. They are attempting to put spyware on our devices to see if we can be sent to jail. That's all that matters. That is the end effect. Apple is justifying putting SPYWARE ON ALL OF THEIR PHONES. Any discussion of the technical merits of a SPYWARE system implemented against you is missing the point. It should not exist.
质疑系统缺乏透明度与问责：哈希库来源不可审计、不可见，还能被远程更新。
这比我最初想的要恶劣得多。他们有一个文件哈希数据库；按设计他们无法验证其内容；他们无法详细说明由谁提供；这些数据的提供者与美国政府合作或关系密切；他们在你自己的设备上，用你自己的文件去比对这个不知道装了什么的数据库；匹配几次后就通知当局；按设计，苹果和你都完全没有可见性——而这对苹果来说恰恰是好事；我猜数据来源可以随时间更新或添加新哈希，而你的设备会乖乖照做。他们唯一关心的是宣称算法如此完美，他们看不到发生了什么，也不会有误报（但愿如此）。行吧，我相信你们能做好，不用多解释。问题在于，你们造的这个东西在设计上就有问题。而苹果显然没有以任何方式回应这一点。但想想孩子们吧……
eknkc查看英文原文
This is a way worse of a spyware than I initially thought. - They have a database of file hashes. - They can’t validate its contents by design. - They can’t explain who supplies it in detail. - The suppliers of these data work with / close to the US gov. - The match your own files on your own device against this database that contains who knows what. - When it matches a couple of times, they alert the authorities. - There is zero fucking visibility for both Apple and you by design and this is a very good thing for Apple. - I assume the source of this data can update / add new hashes in time and your device will happily comply. And their only concern is to say that the algorithm is so perfect, they can not see what’s happening and there won’t be false positives (hopefully). You know what, I trust your ability to do it properly. No need to explain more. Problem is that the very thing you are building is fucked up by design. And obviously Apple does not address it in any way. But think about the children…
质疑安全证明建立在“服务器诚实”和“感知哈希可靠”的假设上，且长期存储的凭证面临量子计算解密风险。
苹果为其系统寻求形式化安全分析是好事，但系统里内嵌了一些根本性的安全假设：(1) 它假设服务器不会篡改数据集（即 CSAM 列表），所以客户端匹配足够多次后披露信息是 OK 的；但现实中没有任何机制阻止恶意服务器往列表里加入任意内容。(2) 它没有考虑感知哈希的脆弱性，包括误报和对抗性碰撞攻击（https://arxiv.org/abs/2011.09473）。另一个长期隐忧是，不清楚苹果会保存 safety voucher 多久；作为存储服务商，苹果可能永久保存。系统基于椭圆曲线密码学，虽是当前最先进的技术，但将来量子计算机成为现实时会被攻破，所以未来 50 年内每一份加密的 safety voucher 都可能被解密。
xucheng查看英文原文
While it is good that Apple seeks formal security analysis over its proposed system, there are some fundamental security assumptions baked in the system: - It assumes that the server would not tamper its dataset (i,e., the list of CSAM). So it is OK to disclose the information if the client has enough matchings. But in reality, nothing prevents a malicious server adding arbitrary content to the list. - It fails to consider the vulnerabilities of the perceptual hash. This includes false positives and adversarial collision attacks (https://arxiv.org/abs/2011.09473). Another potential long term issue is that it is unclear how long Apple will store the safety vouchers. As a storage service, Apple may store them forever. The system is based on Elliptic-curve cryptography. Despite it is the current state-of-the-art encryption technique, it will be broken when the quantum computer becomes a reality in the future. So it is possible that every encrypted safety vouchers can be decrypted in the next 50 years.
质疑隐私设计恰好导致误报无法被人工复核，用户可能在不知情下被报警。
（论文写道：）服务器隐私：恶意客户端除数据集大小外不应获知服务器数据集 X 的任何信息，尤其不能得知交集大小 |id(Y̅ ∩ X)|，否则客户端可以通过往自己的列表 Y̅ 里加测试项、观察交集大小是否变化来提取 X 的信息。→ 如果你的某张照片是误报的哈希碰撞，你根本无从知晓，直到家门被踹开。客户端隐私：恶意服务器除 ftPSI-AD 相对该集合 X 的输出外，不得获知客户端 Y̅ 的任何信息。→ 苹果无法检查哈希匹配是不是误报，因为他们只拿到匹配的哈希，拿不到触发匹配的照片。所以如果你有一堆误报，家门就会被踹开，没有任何人类察觉问题并介入的机会。协议不要求对恶意客户端给出正确输出……他们的协议并不（也无法）抵御这套系统本应防范的那种攻击。
josephcsible查看英文原文
> Privacy for the server: A malicious client should learn nothing about the server’s dataset X ⊆ U other than its size. In particular, it is important that the client learn nothing about the intersection size |id(Y̅ ∩ X)|. Otherwise, the client can use that to extract information about X by adding test items to its list Y̅, and checking if the intersection size changes. If one of your pictures is a false positive hash collision, you'll have no idea until your front door gets broken down. > Privacy for the client: Let X be the server’s input from which pdata is derived. A malicious server must learn nothing about the client’s Y̅ beyond the output of ftPSIAD with respect to this set X. Apple can't check whether a hash match is a false positive or not, because they only get the matching hashes and not the pictures that triggered them. So if you have a bunch of false positives, your front door is getting broken down, with no opportunity for a human to realize the problem and intervene. > The protocol need not provide correct output against a malicious client. That is, the protocol need not prevent a malicious client from causing the server to obtain an incorrect ftPSI-AD output when the protocol terminates. The reason for this is that a malicious client can always choose to hide some of its data from the PSI system in order to cause an undercount of the intersection. Their protocol isn't (and can't be) secure against the one attack that the people that this system is supposed to catch would actually commit. > Moreover, a malicious client that attempts to cause an overcount of the intersection will be detected by mechanisms outside of the cryptographic protocol. This seems eerie to me but I can't put my finger on why.
相关资源
从这篇论文所有 HN 讨论帖的全部评论里整理出来的链接。
- 评论区提到的论文
历次讨论
- 2021Hacker News 讨论帖168 分 · 98 评
同领域还值得读
如何入侵一台已关机的电脑：在 Intel ME 中运行未签名代码
How to hack a turned-off computer, or running unsigned code in Intel ME
这篇 Black Hat 2017 白皮书讨论 Intel Management Engine（ME）的安全问题：如何在 ME 中运行未签名代码，以及攻破处于关机状态的电脑的相关…
安全与密码学2017 · 543 分 · 107 评专业 · 约 40 分钟编辑精选比特币：一种点对点的电子现金系统（2008）
Bitcoin: A Peer-to-Peer Electronic Cash System (2008)
中本聪 2008 年发表的比特币白皮书，提出了无需可信第三方的点对点电子现金系统。
安全与密码学2017 · 536 分 · 394 评入门 · 约 45 分钟编辑精选GPRS 加密算法 GEA-1 的密码分析表明存在故意弱点
Cryptanalysis of GPRS Encryption Algorithms GEA-1 suggest intentional weakness
首次公开了对 GPRS 加密算法 GEA-1 和 GEA-2 的密码分析攻击。
安全与密码学2021 · 487 分 · 107 评专业 · 约 90 分钟编辑精选Signal 消息协议的形式化安全分析
A Formal Security Analysis of the Signal Messaging Protocol
用形式化方法分析 Signal 消息协议（TextSecure）的安全性：作者在多阶段密钥交换模型下，借助自动化证明工具对协议的密钥协商与 Double Ratchet 密钥更新…
安全与密码学2016 · 353 分 · 230 评专业 · 约 120 分钟编辑精选A2：模拟恶意硬件
A2: Analog Malicious Hardware
论文研究如何在芯片设计阶段植入模拟电路形式的恶意硬件后门，使其难以被常规检测发现。
安全与密码学2016 · 315 分 · 94 评专业 · 约 75 分钟编辑精选