@kaepora
"while preserving privacy"
I genuinely think that Ms. Rodriguez sincerely believes this, which is why it's so irresponsible of us to not be honest about the nature of the systems we're creating.
We made a lot of improvements to our ZK architecture, lowering our prover memory requirement from 870GB to 275GB, nearly a 68% improvement.
Getting our system requirements as low as possible allows for the democratization of the decentralized prover network.
On Windows, when something breaks, you're stuck not being able to fix it.
On Linux, when something breaks, you're stuck not being able to fix it, but at least you know why it's broken.
New blog post!
In this one, I go over one of my favorite cryptography papers. This is one of the most surprisingly useful papers I've read, to the point where numerous people use it without even knowing it!
@AlecStapp
The opposition to nuclear power predates the focus on climate change, and opposing nuclear power is a defining aspect of basically every European green party.
A really common misconception about blockchains is that miners "validate" blocks.
Really, every node validates blocks, miners just contribute to choosing the next block to add to the chain. You can't make people accept invalid blocks even if you control all the mining power.
Yeah, so this is what I've been working on for the past month or so.
Basically, the idea is "redo" universally composable security by using state separable proofs as the foundation.
I think the advantage is a simpler framework, that allows using results from standalone security
Oh, you think E2E encryption is good? Well, what if hypothetically 2 million kittens would be killed unless I was able to read all the messages you've sent to your partner? Bet you wouldn't be so hot on encryption then huh buddy
Very happy with this result. In this paper, I looked at threshold Schnorr signatures, Elgamal encryption, polynomial commitments, etc., and then generalized them all under a new notion of what I call "group reconstruction circuits". 1/
@zetalyrae
1990: Nuclear takes too long to build we need stuff now
2000: Nuclear takes too long to build we need stuff now
2010: Nuclear takes too long to build we need stuff now
2020: Nuclear takes too long to build we need stuff now
2022: Nuclear takes too long to build we need stuff now
Image generation models like Stable Diffusion cost approximately $1 million to train. With that money, you could give each American $300 instead. Think about that.
We found a way to mount *remote timing* attacks on *constant-time* cryptographic code running on modern x86 processors. How is that possible? With
#hertzbleed
! Here is how it works (with
@YingchenWang96
).
@TartanLlama
Experienced Programmer:
*deep breath* first of all, what do you mean by string, what do you mean by uppercase, and what do you mean by library?
My hottest take about Cryptography (or ZK Proofs more specifically), is that many things are hard to learn because we don't truly understand them well at all
This week I wrote a basic explainer on a few fundamental topics in cryptography. While this might be useful to people staring out, I think that even advanced cryptographers might also notice something very interesting with this post. Let me know if see it.
I wrote a blog post explaining how Yao's Garbled Circuits protocol works:
This is a little scheme which lets 2 parties compute some arbitrary function of their secret inputs, without revealing those inputs.
I've said this a few times, but the underlying P2P infrastructure of block chains is very neglected.
Also, the P2P layer might serve as an important legal barrier.
I hesitate to recommend books I haven't finished, but I'm about 200 pages into this one and I can highly highly recommend it as a pretty deep introduction to queuing theory as applied to the analysis of computer systems.
Lots of mental patterns are grounded by theory here.
Someone should invent like "Zero-Knowledge" ZK SNARKs, where not only is the proof succinct, but it also doesn't reveal anything else beyond the truth of the statement.
Is anyone working on this?
If you read older cryptography papers, e.g. late 90s, it's interesting how much simpler and to the point they were. Lots of 10 pagers with just one or two results, and minimal introduction.
Machine Learning is a very centralizing force, because of inherent economies of scale and data moats.
Also, it's often deployed as a perverse extension of bureaucracy. You get opaque and heartless decisions with no logic nor recourse behind them.
I wrote an introduction to the technique of state-separable proofs:
I've found this technique to be a much more enjoyable way of writing proofs for game-based security in Cryptography, and it's actually quite a bit easier to use as well.
network protocols not having encryption built in is a good thing actually. The layer-ability it provides prevents the encrypted layer from rigidifying, allowing evolution.
Imagine what standards we would have to be dealing with if TCP had encryption built in back in the day.
First 30 minutes of talk: What is a zero-knowledge proof
Last 5 minutes of talk: How you can use tensor IOPs with algebraic geometry codes to get a linear time prover
Don't get all this hype around anti-cheat software in video games. If you're worried about people cheating in, say, some FPS game, then just make people generate a ZK proof that their frame of input was correctly computed given the state of the game and all of the other players.
"Professor 😥, I need a way to prove security in a composable way for MPC protocols"
"Good news 😉, esteemed Cryptographer Ran Canetti is in town, he'll be able to help you!"
"But Professor 😭, I *am* Ran Canetti!"
A Nintendo emulator is being built with
@NoirLang
.
Dappicom by
@tonk_gg
will allow you to play your favorite Nintendo games and *prove gameplay outcomes on-chain.*
Prove that you beat Zelda without revealing how.
Gamers, read on to learn more ⬇️