Jeff Martin @cuchaz@mastodon.technology

Fantastic analysis of the Beirut explosion from a biomedical engineer who specialized in blast physics and chemistry.

https://www.wired.com/story/tragic-physics-deadly-explosion-beirut/

TL;DR, the explosion was definitely not caused by a weapon, just careless storage of low explosive materials.

@shadowfacts If you're storing gigabytes of data? Sure, I guess it's more work. But users can change passwords as often as they want, right?. If you're storing enough data that this is actually an issue, then pw change would be a great DOS vector, like another poster pointed out.

@shadowfacts Rotation isn't the issue so much. Users change passwords too. So you'll need the new and old values to do the rotation and do a decrypt/encrypt cycle. The trouble with login tokens is you're probably not going to get the old one and the new one at the same time when the user has multiple devices.

@shadowfacts Actually, don't derive the key from the login token. That will make multi-device access impossible.

You just might not have a good story for convenient logins here.

@shadowfacts Ah, ok... that makes more sense. A second salt on the pw will give you a different hash. I guess that works.

The downside I suppose is you'll have to use pw auth to start every new session. ie, no "remember me" option. Unless you use a login token system (which is basically an autogen pw) and derive the key from the autogen pw.

Also, bcrypt is probably fine.... but you can upgrade to Argon2 if you want to add memory-hardness to your hash.

@shadowfacts Oh wait... you mean a second hash of the pw? ie, not the pw hash you're already using for auth?

@shadowfacts Am I missing something here?

I'm assuming your threat model includes an attacker imaging the server storage. Which means they'd get the encrypted data. They'd also get the password hash. Which means they'd also get the decryption key too, right?

So an attacker wouldn't get the password, but they could still access the encrypted storage using the hash.

I'm not sure the salt adds any extra security here, since the pw is ~32 byte hash.

Ps, you're using Argon2, right?

I just typed a super long regular expression, ran it against a file, and it matched perfectly the first time!

Damn, now I feel like a rockstar!

Twitter shares details on the hack.

Looks like hackers wrangled login credentials out of Twitter employees somehow and used internal admin tools to control high-profile accounts:

https://twitter.com/TwitterSupport/status/1283591846464233474

Holy crap! This is huge!

Massive Twitter hack totally breaches trust between the platform and its wealthiest most powerful users:

https://techcrunch.com/2020/07/15/twitter-accounts-hacked-crypto-scam/

@shadowfacts ouch... so cringey!