BRENNEN

Industries / Technology / Software and SaaS / Cybersecurity

Encrypting data

Scrambling data so that only the intended people can read it, and proving it has not been altered, whether it is stored or crossing a network.

7

Milestones achieved, from 1976 to 2024.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1976

    Keys agreed in public

    Before 1976, two parties who wanted to encrypt their messages had to share a secret key in advance, by courier or in person. Whitfield Diffie and Martin Hellman of Stanford showed how two people can agree on a key by exchanging messages over an open channel, so that an eavesdropper who hears everything still cannot work out the key, and proposed public-key systems, in which a key to encrypt can be published while the key to decrypt stays private. Their paper credits Ralph Merkle with an independent, partial solution of a different form.

    ee.stanford.edu

    Credit

    Whitfield Diffie and Martin Hellman; Ralph Merkle

  2. 1977

    A standard cipher, later cracked

    In 1977 the US government adopted the Data Encryption Standard (DES), designed by IBM and modified by the National Security Agency, as the federal standard for protecting unclassified data. It uses 56-bit keys. In July 1998 the Electronic Frontier Foundation's DES Cracker, built for less than $250,000, tried more than 88 billion keys a second and found the key to an RSA Laboratories challenge after 56 hours; the hidden message read "It's time for those 128-, 192-, and 256-bit keys". NIST withdrew the last DES standard, FIPS 46-3, in May 2005.

    LessonA key length that is safe when a standard is written stops being safe as computing gets cheaper.

    eff.orgw2.eff.orgcsrc.nist.gov

    Credit

    IBM and the US National Bureau of Standards, now NIST; the Electronic Frontier Foundation, which built the cracker

  3. 2001

    A cipher chosen in public

    To replace DES, NIST ran an open, three-year competition: 15 candidate designs from researchers in 12 countries, narrowed to five finalists that cryptographers everywhere were invited to attack. In October 2000 it chose Rijndael, by the Belgian cryptographers Joan Daemen and Vincent Rijmen, for the best mix of security, speed and flexibility, and on 26 November 2001 published it as the Advanced Encryption Standard (AES), FIPS 197, with keys of 128, 192 or 256 bits.

    nist.govcsrc.nist.gov

    Credit

    Joan Daemen and Vincent Rijmen; NIST

  4. 2008

    Debian keys made guessable

    On 13 May 2008 the Debian project announced that a change it had made to its own copy of OpenSSL, first uploaded on 17 September 2006, had made the random number generator predictable. Keys for SSH, OpenVPN, DNSSEC and TLS certificates created on affected Debian systems in those 20 months could be guessed. Debian told users to make all such keys again from scratch and to treat DSA keys used on those systems as compromised.

    LessonA key is only as strong as the randomness it was made from, and fixing the code does not fix the keys already made with it.

    lists.debian.org

    Credit

    Luciano Bello, who found the flaw; the Debian Security Team

  5. 2011

    A certificate authority breached

    Browsers trust a list of certificate authorities, organizations that vouch that an encryption key belongs to a given website. In 2011 attackers breached DigiNotar, a Dutch authority, and issued more than 200 fraudulent certificates for more than 20 domains, by DigiNotar's count as reported by Mozilla, and some were used in real attacks. On 2 September 2011 Mozilla removed DigiNotar from its trusted list entirely, saying it was impossible to know how many fraudulent certificates existed. In June 2013 three Google engineers published Certificate Transparency: public, append-only logs of certificates, so that anyone can audit what authorities issue and spot suspect certificates.

    LessonWhen any trusted authority can vouch for any site, what each one issues has to be on public record.

    blog.mozilla.orgrfc-editor.org

    Credit

    Mozilla; Ben Laurie, Adam Langley and Emilia Kasper of Google

  6. 2016

    End-to-end for a billion users

    On 5 April 2016 WhatsApp finished turning on end-to-end encryption, in which only the people in a conversation hold the keys, for all messages, calls, attachments, voice notes and group chats, by default, for more than a billion monthly users. It used the Signal Protocol, and Open Whisper Systems, the protocol's developer, worked with WhatsApp on the change.

    signal.org

    Credit

    Open Whisper Systems; WhatsApp

  7. 2024

    Standards for the quantum era

    On 13 August 2024 NIST published its first three post-quantum standards, methods meant to stay secure against a future quantum computer able to break today's public-key encryption: FIPS 203 (ML-KEM) for agreeing keys, and FIPS 204 (ML-DSA) and FIPS 205 (SLH-DSA) for digital signatures. NIST had assessed 82 algorithms from 25 countries in an effort it dates to a 2016 call for submissions, and told administrators to start using the new standards immediately. In April 2025 OpenSSH, the standard remote-login software, made a hybrid of ML-KEM and an older method its default.

    nist.govopenssh.org

    Credit

    NIST; the teams behind CRYSTALS-Kyber, CRYSTALS-Dilithium and SPHINCS+

Watch for

  1. Two thirds of traffic post-quantum

    Cloudflare says more than 65% of human traffic to its network is now encrypted with post-quantum key agreement. It targets 2029 for full post-quantum security across its products, including post-quantum authentication, the signatures that prove a server's identity, which is not yet in place. The post does not say how the share is measured.

    Achieved whena party other than Cloudflare, such as a browser maker or a university team, reports a similar share of web connections using post-quantum key agreement.

    blog.cloudflare.com

    Claimed by

    Cloudflare, April 2026

  2. Fewer qubits to break curves

    Google Quantum AI says it has designed quantum circuits that would break 256-bit elliptic-curve cryptography, the kind that secures Bitcoin and Ethereum, with fewer than 1,200 logical qubits and 90 million Toffoli gates (a basic quantum operation), which it puts at under 500,000 physical qubits running for a few minutes, about a twentieth of the physical qubits earlier estimates needed. No quantum computer that size exists. Google withheld the circuits and published a zero-knowledge proof instead, a method that lets others check the claim without seeing the attack.

    Achieved whenoutside cryptographers verify the proof, or publish circuits of similar size that others can check.

    research.googlesecurityweek.com

    Claimed by

    Ryan Babbush and Hartmut Neven, Google Quantum AI, March 2026

Open

  1. 01

    Post-quantum defaults, distro by distro

    A table of whether the current releases of about 10 major Linux distributions, among them Debian, Ubuntu, Fedora, Red Hat Enterprise Linux, SUSE, Alpine and Arch, agree a post-quantum hybrid key exchange by default in their packaged OpenSSH, OpenSSL, GnuTLS, NSS, curl, nginx, Apache, Go and Java, and in VPN software such as strongSwan. Each cell rests on a test in which both ends of the connection run on the tester's own machines and the key-exchange method actually agreed is recorded. Rerun the tests whenever packages change.

    The brief
    Credit · unclaimed
  2. 02

    ML-KEM's long tail, tested

    Run the published ML-KEM test sets against the 50 most-downloaded ML-KEM packages on npm, PyPI, crates.io, Maven Central and Go modules that no outside test has yet covered, and publish pass or fail per package and per test set. ML-KEM is the post-quantum key-agreement method NIST standardized as FIPS 203 in 2024. Each failure goes to its maintainers privately before it is published.

    The brief
    Credit · unclaimed

Open 01 · the brief

Post-quantum defaults, distro by distro

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A table of whether the current releases of about 10 major Linux distributions, among them Debian, Ubuntu, Fedora, Red Hat Enterprise Linux, SUSE, Alpine and Arch, agree a post-quantum hybrid key exchange by default in their packaged OpenSSH, OpenSSL, GnuTLS, NSS, curl, nginx, Apache, Go and Java, and in VPN software such as strongSwan. Each cell rests on a test in which both ends of the connection run on the tester's own machines and the key-exchange method actually agreed is recorded. Rerun the tests whenever packages change.

Why it is still open

Upstream projects document their own defaults. OpenSSH has offered post-quantum key agreement by default since version 9.0 in April 2022, made the NIST-based mlkem768x25519-sha256 its default in 10.0 in April 2025, and from 10.1 warns when a connection is not post-quantum, because recorded traffic could be decrypted later. What a user actually gets depends on the package versions and settings each distribution ships. In August 2026 LWN compared how distributions set the choice and found that none yet prefers post-quantum methods in its configuration file, though a new enough OpenSSL may prefer them by its own defaults; the article describes settings, not the method each component actually agrees. A November 2026 search found that and blog posts about single distributions, but no maintained, tested table across them.

openssh.orglwn.net

Built with what exists

Docker or Podman images of each distribution; openssl s_client and s_server, gnutls-cli, ssh -vv, and short Go and Java test programs; tshark to record each handshake; GitHub Actions to rerun the tests.

  1. First step

    Start Debian 13, Ubuntu 26.04 LTS and Fedora containers, and record which key exchange each one's default ssh and openssl clients agree with a default server in the same container network.

  2. Done looks like

    10 distributions by at least 10 components, every cell backed by a saved handshake log; a second person reruns 20 cells on another machine and gets the same result; tests rerun automatically at least monthly for six months.

Open 02 · the brief

ML-KEM's long tail, tested

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

Run the published ML-KEM test sets against the 50 most-downloaded ML-KEM packages on npm, PyPI, crates.io, Maven Central and Go modules that no outside test has yet covered, and publish pass or fail per package and per test set. ML-KEM is the post-quantum key-agreement method NIST standardized as FIPS 203 in 2024. Each failure goes to its maintainers privately before it is published.

Why it is still open

In March 2026 Symbolic Software released Crucible, a test tool built from bugs found in real audits, and ran it on 15 widely deployed implementations. Three of them accepted a malformed public key and one accepted a private key a byte short; the rest passed. In August 2026 a paper ran NIST's official test vectors against three more implementations, and its author notes that nobody outside has yet reproduced it. The Community Cryptography Test Vectors project (CCTV) publishes vectors for specific mistakes, such as comparing ciphertexts with strcmp(), but lists no results per implementation. A November 2026 search found no results for the many smaller packages that applications also install.

symbolic.softwarearxiv.orggithub.com

Built with what exists

Crucible and its harness templates for Rust, Go and C; CCTV's vector files, Google's Wycheproof vectors and NIST's ACVP test vectors; download counts from each package registry; small harness programs for Python, JavaScript and Java, which an AI coding agent such as Claude Code or GitHub Copilot Fusion can draft (see the Note "Effort, not model"); GitHub Actions to rerun on each new release.

  1. First step

    List the ten most-downloaded ML-KEM packages on npm and PyPI and write Crucible harnesses for three of them.

  2. Done looks like

    50 packages run against every test set, with results reproducible from one command; each failure confirmed by a second person and reported to its maintainers; a count of failures fixed within 90 days.