Post by @hipsterelectron@circumstances.run

d@nny disc@

@hipsterelectron@circumstances.run yesterday

the story of the trojan horse is so fascinating to fascists for a few reasons:

it demonstrates to them that accepting anyone else's work into your inner sanctum will lead to your defeat
it's why backdoors take advantage of trust relationships because they don't believe these should be allowed to exist
the walls enchanted by apollo are a convenient analogy to the detente that exists in mathematics when trying to prove something that cannot be proven with the rules they constrain themselves to. the only way through is by cheating

26 Replies

1 Boost

d@nny disc@

@hipsterelectron@circumstances.run yesterday

this is why P < NP hasn't progressed for decades, just like cryptography and parsing theory. everything should have changed after diffie-hellman key exchange

prom™️

@promovicz@chaos.social yesterday

@hipsterelectron "Only our past heroes can see the truth!"... typicall right-wing nonsense.

d@nny disc@

@hipsterelectron@circumstances.run yesterday

@promovicz i realized very recently that public-key cryptography is completely distinct from secret sharing and making identity into a thing that can be stolen is antithetical to security entirely

prom™️

@promovicz@chaos.social yesterday

@hipsterelectron There's a similar argument at a different level, about non-repudiation vs trusting-authenticity. Oftentimes, people don't actually want full non-repudiation, and it is a higher legal ask - so it can be inappropriate.

1+ more replies

d@nny disc@

@hipsterelectron@circumstances.run yesterday

why on earth does anyone call it P = NP in the first place. is it so shocking that anyone can generate a problem instance you won't be able to solve without their knowledge

d@nny disc@

@hipsterelectron@circumstances.run yesterday

this is also why cross-serial productions and multiple parallel stacks aren't formalized because they can't be solved if you can only go left => right. you have to be able to meet in the middle

d@nny disc@

@hipsterelectron@circumstances.run yesterday

i also don't know if i like boolean satisfiability as a basic example of np-completeness because the graph structure of dependencies is so implicit. i forget who the other guy besides cook was involved, was it rabin? not a fan of rabin

~33 more replies

d@nny disc@

@hipsterelectron@circumstances.run yesterday

i guess it's also weird to me to describe complexity theory as if problems have complexities and not their solutions. and in particular diffie-hellman isn't purely reliant on complexity theory; in practice it's also reliant on randomness generation

d@nny disc@

@hipsterelectron@circumstances.run yesterday

like the decision problem formulation of complexity actively removes information from the optimization formulation which is obnoxious to everyone it gets taught to. what if that's because optimality depends upon the model of computation and the entire mode of hierarchical classification of problem definitions is only true for the particular set of assumptions created to codify the turing machine?

d@nny disc@

@hipsterelectron@circumstances.run yesterday

like parsing can definitely be undecidable to solve in my formulation but the answer (if known) can be verified to conform to the grammar specification. and encryption seeks to achieve "effectively random" output—could this too be made undecidable? could something be known but effectively unguessable?

d@nny disc@

@hipsterelectron@circumstances.run yesterday

i think encryption as strictly defined with a statically sized key and a fixed size input and an algorithm with no other inputs can be brute forced by necessity. but one significant result of my parsing method is to completely fail to disambiguate between the cases of streaming (not all data known at once) or fixed input. the double ratchet method generates new keys constantly, but doesn't stretch randomness across the KDF chain, or vary the key length per message. what if these lengths were somehow more fluid? what if there were not a 1:1 invertible AES-GCM operation per message?

d@nny disc@

@hipsterelectron@circumstances.run yesterday

like intermixed with the ciphertext there would be DH handshake information and the parameterization of that switching occurs according to some function specified with reference to previously shared information. because i don't like that you can just try all keys and get the answer. but if you can't determine the message boundaries then you can't really do that. and if you can also send message content out of order then you break the one very strong assumption that my parsing method relies on

d@nny disc@

@hipsterelectron@circumstances.run yesterday

i don't think anyone has really described any good definition of effectively random but undecidable might be cool and insertions + permutations of variable length ranges might make it possible. then again maybe not

d@nny disc@

@hipsterelectron@circumstances.run yesterday

this is kinda how i want my anonymity network to work and i have been thinking it makes sense for encryption and anonymity to be merged. and confusing the ability to determine message boundaries seems to be a shared goal

d@nny disc@

@hipsterelectron@circumstances.run yesterday

having the encoding of something always depend upon some arbitrary chunking decision from the application sounds like a bad idea but compression absolutely requires the structure of the thing to be made known to it to compress better. so applications sending encrypted streams could very well do the same. for encrypting actual user written text then maybe we accept doing some things automatically. but not yet sure

d@nny disc@

@hipsterelectron@circumstances.run yesterday

oh lmao. what if the input text encodes the message chunking somehow so it can be brute forced into any possible string of that length. i think brute forcing AES might already be like this though

d@nny disc@

@hipsterelectron@circumstances.run yesterday

apparently not? https://en.wikipedia.org/wiki/Advanced_Encryption_Standard i don't fully understand how you can recover the key without knowing the message

Advanced Encryption Standard - Wikipedia

d@nny disc@

@hipsterelectron@circumstances.run yesterday

oh nice a fixed block size of 128 bits. well we can fix that

d@nny disc@

@hipsterelectron@circumstances.run yesterday

Each round consists of several processing steps, including one that depends on the encryption key itself.

see this is why djb's whole thing where he whines about encryption steps that depend upon secret information seem so dangerous to me. not only does all of that require the same memory range with the key information to be reused across multiple security contexts in the first place but if you make encryption depend upon something that does not "leak" the key data, then you're reducing the information the key is allowed to transfer to the encryption process

d@nny disc@

@hipsterelectron@circumstances.run yesterday

and like any encryption process obviously depends upon the message itself, which is also secret data??? obviously??? and i know there are stream ciphers djb makes which ostensibly depend upon the message itself. but the only go......LEFT TO RIGHT!!!! i.e. THE SAME PROBLEM WITH EVERY PARSER EVER!!!!

d@nny disc@

@hipsterelectron@circumstances.run yesterday

i also don't get why a "substitution-permutation network" is used when:

substitution is widely understood to not actually obscure any information
permuting the bits within a byte is also a substitution, although the number of unique permutations depends upon the number of set and unset bits (00000000 has only one permutation, but 10101010 has many)

d@nny disc@

@hipsterelectron@circumstances.run yesterday

ohhhhhh wait that's neat. you could argue that compression would recognize this as an opportunity to reduce the number of bits used. so perhaps that would make compression and permutation into ❤️ best friends ❤️

prom™️

@promovicz@chaos.social yesterday

@hipsterelectron how come, encryption doesn't "want" to fiddle with sizes? it would actually make sense from a confidentiality standpoint?

d@nny disc@

@hipsterelectron@circumstances.run yesterday

@promovicz so zstd docs link to this page https://en.wikipedia.org/wiki/CRIME and it's always smelled fishy to me because it relies on compressing the secret key in plaintext alongside the page content. it also requires the page to be publicly accessible to the attacker which like ok then why are you encrypting this at all? and finally like you said if measuring length is a concern then TLS not modifying the length of encrypted content is its own problem!!

CRIME - Wikipedia

prom™️

@promovicz@chaos.social yesterday

@hipsterelectron Under some circumstances you can readily identify some things that an SSH session might be doing, by the size/time patterns of its traffic flow, which goes back to things like ncurses screen updates. It's absolutely guessable in some cases.

d@nny disc@

@hipsterelectron@circumstances.run 22 hr. ago

@promovicz this seems somewhat hard to solve for any latency-sensitive applications which is why i am so harsh on tor since it really needs to be a store-and-forward network