Cryptohaze.com

http://www.youtube.com/watch?v=UHk2lMF7-H4

That is all.

AWESOME, watching now.

Thank you

Feedback welcome.

I honestly can't think of anything that I'd want to improve on, I really liked it. It was easy to follow, the slides did their job of illustrating the scales involved. You covered all your corners with custom definitions and whatnot. It's everything I'd expect from a professional presentation.

I like to see a little bit of Q&A at the end of the talks, usually the Defcon audience do a good job of probing the discussion in an interesting way, that's 100% just personal preference though.

Any plans to return to Defcon and do more talks in future?

frosty wrote:I honestly can't think of anything that I'd want to improve on, I really liked it. It was easy to follow, the slides did their job of illustrating the scales involved. You covered all your corners with custom definitions and whatnot. It's everything I'd expect from a professional presentation.

Thanks. I tried to ensure it was level appropriate for a general audience - I really hate going to talks at Defcon or Blackhat where the whole point of the talk is "Look at how awesome I am, I'm so much smarter than everyone in the room." I don't think it's intentional - it's just that not everyone has the same background.

I like to see a little bit of Q&A at the end of the talks, usually the Defcon audience do a good job of probing the discussion in an interesting way, that's 100% just personal preference though.

Any plans to return to Defcon and do more talks in future?

There was a bit of Q&A, though not as much as I'd like. I aimed for about 45 minutes and hit that pretty accurately.

I'd like to do more talks in the future. I'll be speaking at Passwords^12 in Norway this December, and hopefully doing some additional talks elsewhere. I'd like to find a way to do this whole thing full time.

I hope you do, I find the whole thing fascinating to be honest.

From a systems administrator standpoint it extends past cryptography into the user space, so I'm interested in not just how weaknesses in the Crypto systems put networks and data at risk but also how you go about mitigating that problem, and furthermore how those mitigations change user behavior and how potentially bad user behavior can end up making the system weak for an entirely different reason.

For example implementing strong password requirements such as len10+ with char requirements can cause users to adopt bad practices, write their passwords on postit notes, use common word chains (eg catdogelephant) which could be brute forced with a clever dictionary attack, and I even read a blog about keyboard walking and potentially creating a password list of all possible keyboard walking combinations which is interesting.

As a systems administrator I'm acutely aware that sometimes best practice in theory all goes to shit when given to actual users.

I guess you could say that from a crypto point of view that is solvable by simply moving from performance encryption to deliberately slower encryption which takes more CPU cycles and punishes brute force attacks, but in the real world right now I'm not convinced that's widely adopted.

A talk on all of that would be super interesting.

frosty wrote:I hope you do, I find the whole thing fascinating to be honest.

Thanks! It's a fascinating field.

frosty wrote:From a systems administrator standpoint it extends past cryptography into the user space, so I'm interested in not just how weaknesses in the Crypto systems put networks and data at risk but also how you go about mitigating that problem, and furthermore how those mitigations change user behavior and how potentially bad user behavior can end up making the system weak for an entirely different reason.

I have a few thoughts on this that will become apparent in the next months. I've got some stuff I'd like to get written related to this.

frosty wrote:For example implementing strong password requirements such as len10+ with char requirements can cause users to adopt bad practices, write their passwords on postit notes, use common word chains (eg catdogelephant) which could be brute forced with a clever dictionary attack, and I even read a blog about keyboard walking and potentially creating a password list of all possible keyboard walking combinations which is interesting.

Word chains are an interesting problem to attack, and I have not yet written a dictionary cracker like this, but am seriously considering doing so. Keyboard patterns are total rubbish and are easily pwned by anyone with a moderate dictionary.

frosty wrote: guess you could say that from a crypto point of view that is solvable by simply moving from performance encryption to deliberately slower encryption which takes more CPU cycles and punishes brute force attacks, but in the real world right now I'm not convinced that's widely adopted.

It's not widely implemented, and it gives an attacker an easy way to DOS your site - just try a lot of logins.

frosty wrote:A talk on all of that would be super interesting.

I'll see what I can do!

I'd not considered the stronger crypto stuff being used as a route to DOS the systems using it, a very real possibility, I guess that can at least be mitigated with throttling of login attempts.

Still there has to be a middle ground where you aim for value that servers can cope with en masse but make it too hard to brute force, there has to be some kind of goldylocks zone there. I suppose with hardware speed growing so fast that zone would be constantly moving as well.

It probably wouldn't be that hard to jerryrig a system that simply uses nested encryption on itself X many times where X is dynamic and is correlated in some way to the current date, that way it can scale with computing power.

Register a username/password in 2012 and get a password that's SHA1(SHA1(SHA1(password))), except instead of SHA1 x 3 you have SHA1 x 1000. And in 2014 you get SHA1 x 2000 or something like that. OK, I'm waffling now

Those are still unsalted. Swiss cheese.

But, yes, being able to scale strength with time is an important factor that current algorithms don't really handle.

Yeah you could add salts to them as well I guess, it makes me wonder what sort of clever systems akin to rainbow tables could be built to combat nested encryption like that, and what would happen to systems that heavily rely on it when computing power stops following moores law, a lot to consider. I guess constantly revising your password systems is really the only golden rule here, there seems to be no one fix fits all solution.

It's funny how thinking about how to break these systems is the best way to try thinking about better systems of protection, and how ultimately it's user education that's the best fix for this. That's the messiest solution as far as I can see, some people just don't take it seriously and never will, some of the things I see as a systems administrator not only from users but other sysads is terrifying :s

Anyway, yes, more talks, always more talks! I'm working backwards through Defcon videos to about Defcon 16 now, I have such a massive amount of respect for you guys.