Monday, January 22, 2018

"Skyfall attack" was attention seeking

After the Meltdown/Spectre attacks, somebody created a website promising related "Skyfall/Solace" attacks. They revealed today that it was a "hoax".

It was a bad hoax. It wasn't a clever troll, parody, or commentary. It was childish behavior seeking attention.

For all you hate naming of security vulnerabilities, Meltdown/Spectre was important enough to deserve a name. Sure, from an infosec perspective, it was minor, we just patch and move on. But from an operating-system and CPU design perspective, these things where huge.

Page table isolation to fix Meltdown is a fundamental redesign of the operating system. What you learned in college about how Solaris, Windows, Linux, and BSD were designed is now out-of-date. It's on the same scale of change as address space randomization.

The same is true of Spectre. It changes what capabilities are given to JavaScript (buffers and high resolution timers). It dramatically increases the paranoia we have of running untrusted code from the Internet. We've been cleansing JavaScript of things like buffer-overflows and type confusion errors, now we have to cleanse it of branch prediction issues.

Moreover, not only do we need to change software, we need to change the CPU. No, we won't get rid of branch-prediction and out-of-order execution, but there things that can easily be done to mitigate these attacks. We won't be recalling the billions of CPUs already shipped, and it will take a year before fixed CPUs appear on the market, but it's still an important change. That we fix security through such a massive hardware change is by itself worthy of "names".

Yes, the "naming" of vulnerabilities is annoying. A bunch of vulns named by their creators have disappeared, and we've stopped talking about them. On the other hand, we still talk about Heartbleed and Shellshock, because they were damn important. A decade from now, we'll still be talking about Meltdown/Spectre. Even if they hadn't been named by their creators, we still would've come up with nicknames to talk about them, because CVE numbers are so inconvenient.

Thus, the hoax's mocking of the naming is invalid. It was largely incoherent rambling from somebody who really doesn't understand the importance of these vulns, who uses the hoax to promote themselves.

1 comment:

  1. actually, Solaris had^Whas separate kernel and process address spaces:

    https://books.google.com/books?id=3Ys27a_I1tEC&lpg=PT552&dq=solaris%20internals%20sparc%20address%20space&pg=PT552#v=onepage&q=solaris%20internals%20sparc%20address%20space&f=false

    on certain SPARC microarchitectures

    if i understand what you are getting at--or what Mauro and McDougall
    are talking about, for that matter

    ReplyDelete

Note: Only a member of this blog may post a comment.