(a) informal approach to cryptography, the entire applied cryptography from late 1990s to the early 2000s suffered from this problem, we are still working hard to correct them since Post-Snowden. TLSv1.3 is the most recent effort to pay our debt.
We should understand that, even we can create lots of clever things by mixing and meshing different cryptographic primitives, if the cryptosystem is fundamental enough that you need to use them in serious systems, they must be formally designed and analyzed. Linux’s /dev/random works by gathering various sources of entropy and mixing them, sounds extremely secure, right? But it doesn’t have the rigorous security properties as shown by formal analysis.
One may argue the design is practically secure, but for something fundamental like this, formal approach should be used to keep it robust in all possible and impossible circumstances, just like a good symmetric cipher should resist all forms of known cryptanalysis, to ensure a strong fallback security — that even the adversary can ask you to encrypt any data of their choice, the cipher is still unbreakable.
Empiricism works well in many cases, and you can surely design cryptographic protocols and applications in this way, but it should know when it is needed to stop and call a true cryptographer. The Signal Protocol is a good example of well-designed, and formally analyzed protocol, while Telegram is a bad one.
(b) Overemphasize of Technical Possibilities, But Not Usability. The Cypherpunks were fascinated by cryptography’s endless possibilities. As a result, they create systems which there are hundreds of user adjustable parameters and options.
Do you want AES? TwoFish? 3DES? The first one is the U.S. national standard, well-reviewed, fast hardware implementation, but it has a relatively weak key schedule and may have a small risk of related-key attack in 256-bit mode, but related-key is not a real issue in proper encryption, and… the second one is the security guru Schneier’s cipher, years of cryptanalysis didn’t discover any flaws, and the pure software implementation is typically faster than 256-bit AES, but… Do you want SHA-1? RIPEMD-160? SHA-256? SHA-512? SHA-3? BLAKE2e? SHA-1 and RIPEMD-160 is not recommended for new systems, SHA-256 is the best practice, but why don’t you use SHA-512 for additional protections? By the way, SHA-3 is not vulnerable to length-extension attacks as it’s not a Merkle–Damgård construction, and BLAKE2e is the state-of-art by utilizing Daniel J. Bernstein’s ChaCha20 stream cipher…and so on, and so on, and so on…
You can see this in GnuPG, almost every aspect of encryption can be adjusted. Another example is Mixmaster remailer, which has more than 20 parameters to control the behavior of mail forwarding.
These designs are clearly from the desires to fit the personal preferences by different professional Cypherpunks, but has little actual security benefits. Instead, it greatly complicates the entire protocol, and expose a large surface for attackers. On the other hand, WireGuard, an execellent VPN protocol, doesn’t allow the user to change anything. It is hard-coded to only use state-of-art, and proven algorithms, such as ChaCha20 and Curve25519.
Dreaming about the endless use-cases is another related issue, Cypherpunks spent too much time discussing them, such as trustless key management, etc. However, to bring real changes, we must have systems designed for ordinary people. An less-than-ideal encryption protocol that is simple enough for mass adoption, is superior to a completely trustless protocol, but only usable by a handful of Cypherpunk.
Cypherpunk may disregard these protocols for its imperfections by Cypherpunk’s standard, but it’s still essential to develop them as mass adoption is crucial for Cypherpunks to succeed.
(c) Unrealistic Overconfidence in Cryptography
In Tim May’s The Crypto Anarchist Manifesto of 1992, it announced that, there will be the extensive re-routing of encrypted packets and tamper-proof boxes which implement cryptographic protocols with nearly perfect assurance against any tampering, and as more and more Cypherpunk decided to opt-out from the conventional social order and enters the cypherspace, the State can be therefore eliminated.
The security of any real-life systems is nowhere close to that level in the Manifesto, it could be done in theory doesn’t mean it can be done in practice.
Also, as Cory Doctorow stated, the best encryption, the fastest computers, the most open networks, will not make you comfortable living in an autocratic, corrupt state. You and your radical friends will eventually make a mistake and be rolled up by state thugs, or blacklisted, or blackmailed, or publicly discredited. To evade this all-pervading power, you have to be perfect. To defeat your evasion, the state need merely find a moment’s imperfection in your operational security. Even if your system is perfect, the human-element is the weakest link, the State just need to wait you making a single mistake to be hunt you down, the story of The Silk Road is a good cautionary tale for us.
We should continue the original vision and ideals, but also adopt an reality-based approach, keep quoting Doctorow,
The internet is a tool that can crack open a space in even the most totalitarian of regimes, a place where reformers and revolutionaries can organize, mobilize, and fight back. It’s a forum for whispering dissidence in secret and for blasting the shameful secrets of the powerful at full volume.
The theory of change that goes, “We will walk away from politics and use the internet to evade state oppression” is a dead letter. It always has been. But the theory that goes, “The internet will let us organize to hold the government to account, to topple the corrupt, to rally the honorable and expose the wicked” — that theory has never been more important.
(d) “Cypherpunks Write Code!”
Linus Torvalds like to say “Talk is cheap, show me the code”. The slogan of Cypherpunks in fact came earlier, “Cypherpunks Write Code!”. The meaning is that, good ideas need real implementation, not just talks.
Unfortunately, the discussions of the Cypherpunks were too broad and pretty much ahead of its time, as a result, most ideas were not implemented, and few implementations were only prototypes. Even the best software package, suffers from Problem B.
It is not to say that discussions are unnecessary, but we must build something for mass adoption as well. So perhaps a new slogan, “Cypherpunks Build Apps” can be used, although “app” sounds like a corporate buzzword.
(e) Hostile Discussion Atmosphere
Cypherpunks were radicals, many are radical individualists and anarcho-capitalists, with very strong personally opinions on almost everything and very militant. The Cypherpunk mailing list suffered from endless personal attacks and arguments from day 0. I recommend everyone to read the Cypherpunk Mailing List archives to understand the situations.
If we want to build a public forum, we should take all the possible precautionary measures to prevent it from happening, and cooperate towards a common goal. Having an ideological civil war between democratic socialist, and libertarians only helps the Big Brother.
We should find a way to allow diverse groups of people of different ideologies to cooperate to accomplish things, meanwhile still upholding the same core values.