Tuesday, June 12, 2018

Quotes from Dan Kaminsky's Keynote at DEF CON China


Above is Dan Kaminsky's keynote at the inaugural DEF CON China.  It was nominally about Spectre and Meltdown, and I thought it was immediately applicable to testing at all levels.  Here are some moments that jumped out at me:

On Context:

"There's a problem where we talk about hacking in terms of only software...What does hacking look like when it has nothing to do with software." 1:55

"But let's keep digging." Throughout, but especially 5:40

"Actual physics encourages 60 frames per second. I did not expect to find anything close to this when I started digging into the number 60...This might be correct, this might not be. And that is a part of hacking too." 6:10

"Stay intellectually honest as go through these deep dives. Understand really you are operating from ignorance. That's actually your strong point. You don't know why the thing is doing what it is doing...Have some humility as you explore, but also explore." 7:40

"We really really do not like having microprocessor flaws...and so we make sure where the right bits come in, the right bits come out. Time has not been part of the equation...Security [re: Specter/Meltdown] has been made to depend on an undefined element. Context matters." 15:00

"Are two computers doing the same thing?...There is not a right answer to that. There is no one context. A huge amount of what we do in hacking...is we play contexts of one another." 17:50

[Re: Spectre and Meltdown] "These attackers changed time which in this context is not defined to exist...Fast and slow...means nothing to the chip but it means everything to the users, to the administrators, to the security models..." 21:00

"Look for things people think don't matter. Look for the flawed assumptions...between how people think the system works and how it actually does." 35:00

"People think bug finding is purely a technical task. It is not because you are playing with people's assumptions...Understand the source and you'll find the destination." 37:05

"Our hardest problems in Security require alignment between how we build systems, and how we verify them. And our best solutions in technology require understanding the past, how we got here." 59:50

On Faulty Assumptions:

"[Example of clocks running slow because power was not 60Hz] You could get cheap, and just use whatever is coming out of the wall, and assume it will never change. Just because you can doesn't mean you should...We'll just get it from the upstream." 4:15

"[Re: Spectre and Meltdown] We turned a stability boundary into a security boundary and hoped it would work. Spoiler alert: it did not work." 18:40

"We hope the design of our interesting architectures mean when we switch from one context to another, nothing is left over...[but] if you want two security domains, get two computers. You can do that. Computers are small now. [Extensive geeking out about tiny computers]" 23:10

"[RIM] made a really compelling argument that the iPhone was totally impossible, and their argument was incredibly compelling until the moment that Steve Jobs dropped an iPhone on the table..." 25:50

"If you don't care if your work affects the [other people working on the system], you're going to crash." 37:30

"What happens when you define your constraints incorrectly?... Vulnerabilities. ...At best, you get the wrong answer. Most commonly, you get undefined behavior which in the presence of hacking becomes redefinable behavior." 41:35

"It's important to realize that we are loosening the assumption that the developer knows what the system is supposed to do...Everyone who touches the computer is a little bit ignorant." 45:20

On Heuristics

"When you say the same thing, but you say it in a different time, sometimes you're not saying the same thing." 9:10

"Hackers are actually pretty well-behaved. When hackers crash code...it does really controlled things...changing smaller things from the computer's perspective that are bigger things from a human's perspective." 20:25

"Bugs aren't random because their sources aren't random." 35:25

"Hackers aren't modeling code...hackers are modeling the developers and thinking, 'What did [they] screw up?' [I would ask a team to] tell me how you think your system works...I would listen to what they didn't talk about. That was always where my first bugs came from." 35:45

On Bug Advocacy

"In twenty years...I have never seen stupid moralization fix anything...We're engineers. Sometimes things are going to fail." 10:30

"We have patched everything in case there's a security boundary. That doesn't actually mean there's a security boundary." 28:10

"Build your boundaries to what the actual security model is...Security that doesn't care about the rest of IT, is security that grows increasingly irrelevant." 33:20

"We're not, as hackers, able to break things. We're able to redefine them so they can't be broken in the first place." 59:25

On Automation

"The theorem provers didn't fail when they showed no leakage of information between contexts because the right bits went to the right places They just weren't being asked to prove these particular elements." 18:25

"All of our tools are incomplete. All of our tools are blind" 46:20

"Having kind of a fakey root environment seems weird, but it's kind of what we're doing with VMs, it's what we're doing with containers." 53:20

On Testing in the SDLC

"We do have cultural elements that block the integration of forward and reverse [engineering], and the primary thing we seem to do wrong is that we have aggressively separated development and testing, and it's biting us." 38:20

"[Re Penetration Testing]: Testing is the important part of that phrase. We are a specific branch of testers that gets on cooler stages...Testing shouldn't be split off, but it kinda has been." 38:50

Ctd. "Testing shouldn't be split off, but it kinda has to have been because people, when they write code, tend to see that code for what it's supposed to be. And as a tester, you're trying to see it for what it really is. These are two different things." 39:05

"[D]evelopers, who already have a problem psychologically of only seeing what their code is supposed do, are also isolated from all the software that would tell them [otherwise]. Anything that's too testy goes to the test people." 39:30

"[Re: PyAnnotate by @Dropbox] 'This is the thing you don't do. Only the developer is allowed to touch the code.' That is an unnecessary constraint." 43:25

"If I'm using an open source platform, why can't I see the source every time something crashes? ...show me the source code that's crashing...It's lovely." 47:20

"We should not be separating Development and Testing... Computers are capable of magic, and we're just trying to make them our magic..." 59:35

Misc

"Branch Prediction: because we didn't have the words Machine Learning yet. Prediction and learning, of course they're linked. Kind of obvious in retrospect." 27:55

"Usually when you give people who are just learning computing root access, the first thing they do is totally destroy their computer." 53:40 #DontHaveKids

"You can have a talent bar for users (N.B.: sliding scale of computer capability) or you can make it really easy to fix stuff." 55:10 #HelpDesk
"[Re: Ransomware] Why is it possible to have all our data deleted all at once? Who is this a feature for?!... We have too many people able to break stuff." 58:25
Post a Comment