12 January, 2018

Nasty fall-out from Spectre and Meltdown

I guess it's hard to miss Spectre and Meltdown so you probably read about it. And there's more bad news than what's been widely reported, it seems.

You trust the cloud? HAHAHAHA

What surprised me a little was how few journalists paid attention to the fact that Meltdown in particular breaks the isolation between containers and Virtual Machines - making it quite dangerous to run your code in places like Amazon S3. Meltdown means: anything you have ran on Amazon S3 or competing clouds from Google and Microsoft has been exposed to other code running on the same systems.

And storage isn't per-se safe, as the systems handling the storage just might also be used for running apps from other customers - who then thus could have gotten at that data. I wrote a bit more about this in an opinion post for Nextcloud.

We don't know if any breaches happened, of course. We also don't know that they didn't.

That's one of my main issues with the big public cloud providers: we KNOW they hide breaches from us. All the time. For YEARS. Yahoo did particularly nasty, but was it really such an outlier? Uber hid data stolen from 57 million users for a year, which came out just November last year.

Particularly annoying if you're legally obliged to report security breaches to the users it has affected, or to your government. Which is, by the way, the case in more and more countries. You effectively can't do that if you put any data in a public cloud...

Considering the sales of the maximum allowed amount of stock just last November by the Intel CEO, forgive me if I have little trust in the ethical standards at that company, or any other for that matter. (oh, and if you thought the selling of the stock by the Intel CEO is just typical stuff, nah, it was noticed as interesting BEFORE Meltdown & Spectre became public)

So no, there's no reason to trust these guys (and girls) on their blue, brown, green or black eyes. None whatsoever.

Vendors screwed up a fair bit. More to come?

But there's more. GregKH, the inofficial number two in Linux kernel development, blogged about what-to-do wrt Meltdown/Spectre and he shared an interesting nugget of information:
We had no real information on exactly what the Spectre problem was at all
Wait. What? So the guys who had to fix the infrastructure for EVERY public and private cloud and home computer and everything else out there had... no... idea?

Yeap. Golem.de notes (in German) that the coordination around Meltdown didn't take place over the usual closed kernel security mailing list, but instead distributions created their own patches. The cleanup of the resulting mess is ongoing and might take a few more weeks. Oh, and some issues regarding Meltdown & Spectre might not be fix-able at all.

But I'm mostly curious to find out what went wrong in the communication that resulted in the folks who were supposed to write the code to protect us didn't know what the problem was. Because that just seems a little crazy to me. just a little.