bren2013

Are you interested in particular types of scientist? I.e., physicists, chemists, biologists? Computer scientists? :)

> I don't know of any issuer key compromises detected with CT. Such compromises are rare. DigiNotar is an obvious example, but can you think of any recent ones?

Not key compromise, but general mis-issuance:

Facebook detected overreach by a vendor with CT: https://www.facebook.com/notes/protect-the-graph/early-impac...

AGL detected certs that were malformed in various ways: https://www.imperialviolet.org/2013/08/01/ctpilot.html

> Technically CT is still not required.

It produces an interstitial, see: https://invalid-expected-sct.badssl.com/

Is Cirrus mentioned in the blog post? I can't find it.

Right now, the way key compromise might be detected in RPKI is that a human network operator notices a signed object which is obviously suspicious and posts it on a mailing list. This is the same way that CA compromise was detected in the Web PKI before CT.

CT was useful well before Chrome used their weight to make it required for all new certificates because it was no particular burden on a few people (CA or not) who saw a lot of certificates to add them to CT. This gave the people who might not see a lot of certificates but want to find something weird a corpus to dig into.

Same idea applies to RPKI. But yes, setting up Cirrus took very little of my time.

IPFS offers content integrity, which you can use to bootstrap confidentiality.

Not sure if this fully addresses your use-case, but I like the idea of serving a static bootloader from IPFS. The bootloader would contain all of a website's assets, and code for getting dynamic content from a backend. The backend could be:

- A central API where the bootloader can do arbitrary validation on the API responses.

- WebTorrent, Scuttlebutt, IPFS PubSub, etc.

I believe the IPFS team has already built a browser extension that runs a js-ipfs node. Running an IPFS node is not ideal for power and bandwidth constrained devices like a mobile phone. This is where I see gateways fitting in, long-term. Not having to trust a gateway unconditionally is necessary to make this viable.

No plans currently. I'm not really familiar with Dat. Can it do something that IPFS can't?

Kiwix has an archive of StackOverflow but its gigantic. :( We barely got the Math StackExchange uploaded in time.

> And ultimately, how can I know for sure that CloudFlare won't play the game where as acting as a proxy they will modify some of the files served?

https://blog.cloudflare.com/e2e-integrity/

There are lots of scenarios where you would want to let an untrusted intermediary handle critical operations for you without actually having to build trust with them.

Typical example: Certificate authorities distribute CRLs primarily through untrusted channels and third parties to lower their operating cost. The CRLs are signed so its a win-win for everybody.

Not for a while. DJB's curves have entirely too much structure and I'd prefer to get something general purpose working first.

I don't even like having to implement NIST's speedup for A = -3, and it's fairly trivial.

I've tested a scalar of about 50% 1s against a scalar with almost no 1s, measuring at the nanosecond resolution, with 100 samples each.

The p-value of the 2-sample T-test was greater than 1%--there's no evidence that one takes less time than the other. That can be replicated by playing with the benchmarks yourself. If you have any evidence to the contrary, I'd be happy to look at it.

Issue #1: Yes, it's supposed to be like that. The point is that any n-bit scalar takes the same amount of time as any other n-bit scalar.

Issue #2: Rust has explicitly taken all memory management away from me. There's nothing I can do about that.

I missed my edit window...

I actually take the above back. Scalar multiplication on a point IS constant-time (the two distributions are indistinguishable).

Field exponentiation isn't constant-time and I will work on that.

About ElGamal: Agreed. I only planned to implement it for my own satisfaction, more than for anything else.

The addition law only leaks general cases--doubling, negatives, one argument is zero. Is that not generally considered okay? Intuitively, it doesn't seem like terribly valuable information--in the case of the Montgomery ladder, the attacker already knows we do one doubling and one addition per bit.

Earlier, I was using Euler's theorem for inversions, which should have been more constant-time than Euclid's algorithm. The only problem is, it took slightly over a minute to do one scalar multiplication. Could you describe an attack that would come from knowing information about the z-coordinate? Is there a good way to do constant-time inversion?

Edit: I just added a second benchmark testing scalar multiplication for a random value vs one with lots of zeros, and they produce different distributions with or without the #[const_time] (assuming I'm using it right). Thank you for bringing this up! I'll look into what needs tweaking.

Uber employees are the kind of people who kept a telescope in their bedroom window to peep on girls down the street. I'll never understand why they're still in business.

Generally, when you backdoor hardware, you don't let people know it's backdoor'ed.