Encrypting mails with usable tools
The mass adoption of encryption technologies
Kali Kaneko
Encryption is the application of mathematics to ensure that our information is only accessible to the persons or machines we decide to share it with. Encryption has a long history. Protocols for sending encrypted information without having previously agreed on a shared secret with the other party (for deciphering encrypted data) have been around for roughly 35 years. The landmark Pretty Good Privacy program, often abbreviated as PGP, made strong email encryption with guarantees of confidentiality, authentication and integrity widely available to users and developers in 1991 0.
The relative popularity of PGP and its subsequent standardization is often depicted as a victory for the cypherpunks (cryptography activists) during the First Crypto Wars 1. So, what are the issues that still prevent the adoption of email encryption by the critical and large masses? Why couldn’t Greenwald, the respected journalist, encrypt an email when he was contacted by his source Snowden, the analyst working for the NSA 2?
To answer that, we have to look first at the architecture of internet services, then the economy of surveillance, and finally some historical usability failures.
Email in the times of surveillance capitalism
Email is an open, federated protocol that has been re-centralized by big service providers. These companies exploit economies of scale that lead to the commoditization of email. Negative externalities from the commoditization of email include the arms race that evolved to prevent the spam industry from sending a large amount of unsolicited and often fraudulent email to users.
In the early days of the internet, anybody could run an email provider. The past decade, however, has seen a drastic reduction in the number of email providers. Not only few individuals and collectives run their own email servers, but less and less people know that it is even possible to do so. Email has become another example of technology that is “just supposed to be there”. The message seems to be that vital infrastructure is not something that you run for fun. Something similar is happening with Instant Messaging 3.
Losing the battle for open, federated architectures also means losing control over the communication infrastructures we use. The increase in monopolistic practices leads to a lack of interoperability between providers that puts up a barrier of entry for new email service providers.
Monoculture is an imperative for centralized control: it means that an adversary only needs the cooperation of a single player to compromise the private communications of millions. Email is not only about messages: today it is one of the last technological lines of defence on the internet for privacy-aware and localised alternatives for Identity Provision. Legitimate email service providers can still choose to allow their users anonymity or pseudonimity, refuse to track or sell their data, etc.
We increasingly see how mobile devices, instead of email accounts, are required for bootstrapping communication with your contacts. The phone has become the gateway to Facebook's walled gardens in many countries 4. The lawyer and privacy advocate Lawrence Lessig warned that the Big Regulatory Loop is closing between the Industry and the State 5, a big feat for those “weary giants of flesh and steel” that we didn't use to fear.
Governments and corporations race towards the deployment of pervasive surveillance. When big powers invest so much effort into eroding the fundamental rights of individuals and communities to decide upon the scope of their communications, building - and using - privacy tools becomes a moral imperative. The right to whisper is an irrevocable and fundamental right that is being taken away by force. Its criminalisation and disappearance has a profound impact on our capacity to exert our human rights and shape democratic societies.
We failed to empower people to encrypt
Strategically, email might seem an odd choice nowadays, in an ever changing technological landscape that leans more and more towards mobile applications, and where most users had their first experience on the internet already mediated by the Big Silos. Email is often criticized as an obsolete technology, because its architecture makes it hard to encrypt messages in a way that hides who is writing to whom about what.
Even if more interesting and attack-resistant technologies appear in the future 6, email will stay around at least for a while. Email is the asynchronous medium we have and that needs to be protected. Millions of emails are still sent daily without encryption, and emails with unverified senders are still used for devastating phishing or ransomware attacks.
It is hard to assess whether it is only a fundamental usability problem that we're dealing with, or if on the contrary we are facing a general loss of interest in: 1. email as a tool; 2. privacy and security in online communications; or 3. a combination of the above.
Commercial-level storage is cheap, and providers that rely on the exploitation of silo data and metadata can afford to give it away “for free”. It’s obvious that these providers do not have a real incentive for encryption, because it would undermine their revenue streams. Even providers that support encryption profit from traffic analysis.
It has been shown that users are willing to pay an extra cost for services or apps that respect their privacy 7, yet privacy-aware email providers have to compete against the major market players who exploit economies of scale to offer a baseline of the 15GB of “free” storage, high standards of reliability, speed, etc. In other words, there are many critical users who could contribute to the costs of privacy-aware email providers, but usability and operational costs add up to making it very hard to compete against the established monopolies.
In this sense, any serious attempt to provide alternatives must address the sustainability of technopolitical projects such as privacy-aware email providers.
It's also a problem with tools for nerds
The “scratch your itch” attitude of the Free Software community just doesn't cut it any more for mass adoption. Self-discipline and quality are key for maintaining a sustainable community around pleasant, usable and effective software.
Given the humongous amount of resources that Capital has invested in the cybernetic control of the masses, current user expectations about usability are high. Interacting with digital tools should only require a very small amount of cognitive effort. New technologies that defy too many conventions (visual language and best practices such as common metaphors, established interfaces, features such as multi-device syncing, mobile first design, etc.) erect barriers against their adoption. The eternal request from users to developers for the well known “wall” or “like” features in new technologies shows that arbitrary signs have became normalized.
Nonetheless, oversimplification at the cost of irrevocably hiding complexity from the user is not the only available option – and it often backfires. One could dream of an interface that makes everyday tasks simple, but allows users to explore further possibilities as they learn more.
Nerd-driven development also shifts focus whenever a new shiny technology emerges. That might partly explain why some technology stacks just languish and rot. We need to cultivate excellence even in pieces of technology that don't get the excitement associated with new developments. If we want crypto to spread outside the techie ghetto, half-baked and unmaintained tools that are unusable should be abandoned. Arrogance about what users need to know or do before accomplishing the most basic task must be held under control.
One example of broken tech is the Web of Trust system used for identifying the keys associated with a given email address. Didactic attempts to explain its necessity for safe encryption practices have failed along the years, perhaps because it was based on broken assumptions from its inception 8.
Looking at possible solutions
Many projects have blossomed in the post-Snowden era. I mention here one I am contributing to, and others that I consider interesting, both in terms of working software and evolving protocols. My focus is on initiatives that build interoperable solutions on the top of the existing email infrastructure using the OpenPGP standard 15. I also briefly mention some new silos that try to monetize on the crypto fuzz.
Bitmask and the LEAP Encryption Access Project
LEAP aims to develop encrypted email services that are easy to deploy and clients that are simple to use 9. LEAP implements opportunistic email encryption, which is a transparent process that requires only a little cognitive effort from users, and low maintenance costs for providers. LEAP software may enable many federated providers to enter the email provisioning space by lowering the technical and economical costs.
On the server side, the LEAP Platform is a set of complementary software packages and recipes to automate the maintenance of LEAP services. Its goal is to make it as painless as possible for sysadmins to deploy and maintain secure communication services, as well as to help providers manage registration and billing.
On the client side, the Bitmask application runs in the background. It acts as a proxy for the same email programs that users are already familiar with. Alternatively, another interface is available that runs in the browser (through a customised version of Pixelated 16. Bitmask finds the relevant encryption keys for email addresses automatically, and works across different devices. All data (including the encryption key database and the email itself) is end-to-end encrypted, which means that service providers have no access to the contents. As part of the Panoramix project 17, anonymous routing capabilities defending against traffic analysis will be added too, providing a greater level of privacy.
Throw your metadata through the Memory Hole
In an email, the data is the content: the letter that you write. The metadata is everything that helps the content be routed to its intended recipient: it is equivalent to the stamp, the envelope and the addresses of the recipient and the sender in a traditional mail.
Conventional email encryption technologies are only concerned with protecting the content of the message. Therefore, the metadata remains visible in transit. Intermediaries who act as postmen can see your address, the recipient, date, subject line and even the path the message took to its destination.
The Memory Hole project aims to fix this problem by stashing metadata in the contents of the e-mail in a standardized way. This means to hide as much metadata as possible inside the “protected” envelope from intermediaries such as service providers or spy agencies.
By implementing this proposed standard, Memory Hole compliant email programs can protect a good amount of metadata from snooping and modification in transit. Look for this feature in the near future!
Autocrypt: Such crypto, much mail
The Autocrypt project develops email encryption that is convenient enough for mass adoption even if it cannot be as secure as traditional email encryption.
The project is driven by a diverse group of mail app developers, hackers and researchers who are willing to take fresh approaches, learn from past mistakes, and thereby increase the overall adoption of encrypted email. Some popular software such as K9 (mobile email app), Enigmail (encryption plugin for the Thunderbird mail reader) or Mailpile (a web interface for email) already support this protocol.
Autocrypt uses regular email messages to exchange the information that allows the encryption of subsequent messages. It adds metadata to the email that stores the encryption keys associated with users, as well as their relevant preferences about encryption behaviour.
The Webmail family: Modern email clients built on web technologies
A webmail interface offers an intuitive user experience. It runs in the browser that is available on any desktop computer. In-browser apps pose some security problems (unverifiable code execution, secrets storage open to a very wide attack surface, etc.) but it also drastically lowers the barrier to adoption.
Mailpile 10 is a self-hosted email service. Its user interface takes advantage of widely supported web standards such as HTML5 and JavaScript. The interface connects to a backend that typically lives on the local device, but may also run on a server. It supports end-to-end encryption via the traditional OpenPGP standard. The interface emphasises searching and tagging, which makes it a bit similar to the popular Gmail web inferface and sets it apart from most other free software email programs. The Mailpile initiative holds a lot of promise as a modern cross-platform mail client, especially since the Mozilla Foundation stopped supporting the development of its main alternative, the Thunderbird desktop mail reader.
Another interesting open-system webmail approach was Whiteout, which closed in 2015 with more than 10.000 users. Their open source software implemented interoperable protocols. In their post-mortem note they shared some calculations about what a viable market of encrypted messaging apps could look like 11, although the lesson might well be that the model of startup companies is not suitable for tackling the surveillance problem.
Mailvelope 12 might be a suitable option when compromises have to be absolutely made. It is a browser extension that allows you to use OpenPGP email encryption with mainstream webmail providers like Gmail, Yahoo, and Outlook. Popular free software webmail applications like Roundcube (the webmail offered by projects like Riseup or Autistici) also support the Mailvelope plugin 13.
Non-email messaging services
Last, I’d like to mention several initiatives that are gaining traction among users that seek secure email providers, but that cannot be considered as interoperable and encrypted email services. They usually support end-to-end encryption only between users of the same service, and fall back to unencrypted email for users of other providers. Alternatively, some require that users across providers exchange a shared secret manually – which defeats the purpose of public key cryptography that is the big thing about the OpenPGP standard used by the other projects, and centralizes the ecosystem once again.
Known examples of this category of centralized, non-email services are ProtonMail (a Swiss webmail app that only does encryption between its own users, sending unencrypted email to others) and Tutanota (a webmail interface and mobile app that requires external recipients to decrypt the message on the Tutanota website with a pre-shared secret).
For in-depth reviews of other initiatives, and a nice overview of projects that support email encryption, an extensive comparison is available online 14.
Some remaining challenges
The quest for reducing the interception of our global communications is still ongoing. The challenge is to collectively recover control over the email medium. As shown above, some projects are making good progress. They are adopting new strategies for achieving mass adoption of easy-to-use email encryption.
The promise is that over the coming months better programs for email encryption can work together in a mostly automated way, demanding less user intervention while still ensuring that users can decide who can see their messages while they travel across the internet.
But programs do not get written alone: I encourage you, especially, to try clients like K9, Enigmail, Mailpile and Bitmask. Test them out. Try more than one. Try them with your friends, with your family. Engage with their communities, join their mailing lists and IRC channels. Learn more about their strengths and limitations. Report problems when they break, try new versions, write or improve translations to your own language, start hosting a new email provider if you can and above all, continue contributing to the process of collective creation. If you believe in the right to whisper, engage in the global conversation and raise your voice.
Take care! I look forward to reading you securely in the intertubes.
0. There are several different properties that crytographic encryption solutions traditionally aspire to provide. Confidentiality is obtained by encrypting messages, which in plain English means scrabbling them in order to avoid third parties (like a government, corporation or malicious person) to recover the content and read them. Authentication is done by signing the content on one end and verifying the signatures on the other end to make sure that the message was really sent by the claimed author. The way the encryption is done also allows to preserve content integrity, ensuring that no third party could change the message in transit. ↩
1. In the Cold War, the USA and its allies developed an elaborate series of export control regulations designed to prevent a wide range of Western technology from falling into the hands of others, particularly the Soviet bloc. Export controls on encryption became a matter of public debate with the introduction of the personal computer. Zimmermann’s PGP and its distribution on the internet in 1991 was considered the first major ‘individual level’ challenge to cryptography export controls, although ultimately, the popularization of e-commerce probably did play a much bigger role in the outcome. https://en.wikipedia.org/wiki/Crypto_Wars ↩
2. When Snowden first tried to contact the Guardian journalist Glenn Greenwald, cryptography hackers and privacy activists collectively experienced a harsh reality check that punched our little Web of Bubble: no security is effective without usability. If an NSA analyst is forced to craft awful videos in order to teach a journalist how to install a tool called gpg4win, downloaded from an ugly website, do some scary copy/pastes and other such delights (shown in the 12 minute video: http://www.dailymail.co.uk/embed/video/1094895.html ), we can strongly conclude that the usability, and general state of email encryption is terribly broken. So, more than ten years after the seminal article, we can affirm that sadly, Johnny cannot yet encrypt: https://people.eecs.berkeley.edu/~tygar/papers/Why_Johnny_Cant_Encrypt/OReilly.pdf ↩
3. In other words: the long death of Jabber/XMPP. It’s frustrating how, over and over again, the fragmentation of an open ecosystem leads to centralized solutions. One can understand Signal developer and crypto anarchist Moxie Marlinspike’s rants against federation only in terms of the desire of deploying updates to millions of users without waiting for the long tail and the distributed consensus to catch up. In the mobile messaging world Signal is right now the best thing we have, but it still represents a failure of the technosocial processes that prevented the open federation of communication infrastructures from becoming a reality today. ↩
4. And with the phone, the policy of mandatory real-name registration. This practice is enforced by telecommuncations companies on behalf of states that pass anti-anonymity laws. ↩
5. http://codev2.cc/download%2Bremix/Lessig-Codev2.pdf ↩
6. Projects like Pond, Retroshare or Secushare might be good insights into what a post-email secure, distributed standard might look like. https://github.com/agl/pond • http://retroshare.us/ • http://secushare.org/ • Note that the Pond author recommends using the Signal app for practical purposes until his own software is more polished and reviewed. ↩
7. See, for instance The Value of Online Privacy and What is Privacy Worth?: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2341311 • https://www.cmu.edu/dietrich/sds/docs/loewenstein/WhatPrivacyWorth.pdf ↩
8. https://github.com/micahflee/trollwot ↩
9. https://downloads.leap.se/publications/cans2016.pdf ↩
10. https://mailpile.is ↩
11. https://tankredhase.com/2015/12/01/whiteout-post-mortem/index.html ↩
12. https://mailvelope.com ↩
13. https://roundcube.net/news/2016/05/22/roundcube-webmail-1.2.0-released ↩
14. https://github.com/OpenTechFund/secure-email ↩
15. https://openpgp.org/ ↩
16. https://pixelated-project.org/ ↩
17. https://panoramix-project.eu/ ↩
18. https://modernpgp.org/ ↩
19. https://autocrypt.readthedocs.io/ ↩