The DMZ

Scott Pack

00:00

Ahoy

Iszi

@ScottPack Yoha

Lucas Kauffman

00:37

Yar har fiddleteee

being a security analyst is alright to be!

do what you want 'cause an analyst is free!

xce

04:50

so you can find a collision for MD5(x) but how about MD5(MD5(x))?

Iszi

05:05

@xce Logically, I don't see why the latter should be any more difficult than the former.

Terry Chia

05:37

@xce I agree with what Iszi said.

xce

I don't get it

Terry Chia

The odds of finding a collision of MD5(x) should be the same as MD5(MD5(x)) simply due to the nature of a hashing algorithm, which maps an input of any length into a fix ed length hash.

xce

there are special techniques to find MD5 collisions quickly though

Terry Chia

There is no reason why said techniques shouldn't work on MD5(MD5(x)). Which is exactly why no one suggest using MD5 anymore.

xce

but can you find a collision in practice?

I couldn't

Terry Chia

05:44

Read up on the Flame malware, a good example of a recent threat that utilizes MD5 collisions.

Iszi

@xce I'm not intimately familiar with the MD5 algorithm, or techniques for finding collisions in any hashing algorithm. However, it would stand to reason that finding a collision for MD5(MD5(x)) should be equally as difficult as finding a collision for MD5(x) itself.

Really, both examples are just different instances of MD5(x).

I think you may be mistaking the way the problem should be approached. To find a collision for MD5(MD5(x)), you would not first need to find a collision for MD5(x). Instead, you would just calculate MD5(x) itself, and then use that as the input to find a collision for what becomes the new MD5(x) problem - where MD5(x) is simply the new x.

Where's @ThomasPornin when you need him...?

xce

06:09

so given x can you find y such that MD5(MD5(x)) = MD5(MD5(y)) that way?

Terry Chia

What Iszi is saying is that you can think of MD5(x)=y, and find the collision for MD5(y)

xce

well I don't really understand

Terry Chia

MD5(MD5(x)) is basically the same as MD5(x)=y, MD5(y)

Trying coding it out to visualize it better.

xce

def win(x, y):
    return (x != y && md5(md5(x)) == md5(md5(y)))

try to make this function return true basically

Terry Chia

What language is that?

xce

06:16

python

Terry Chia

Ah, I'm not familiar with C.

xce

can't edit any more && should be and

makes sense?

Terry Chia

In pseudo code,

hash = MD5(MD5(x));

will return the same value as

y = MD5(x);
hash = MD5(y);

xce

oh yeah good point

Terry Chia

So to find a collision for MD5(y), you would not have to find the collision of MD5(x) before finding the collision of MD5(MD5(x)). You could simply find the collision for MD5(y).

As for the specific details on how to go about doing it, I am not too sure.

xce

06:21

cool idea thanks

Terry Chia

If you really still cannot understand the specifics, I suggest asking a question on crypto.SE. I'm sure experts over there could give a far better and more in-depth answer.

xce

I might post it there

although I'll wait because I have an unanswered question there already

I need a proper book rather than just google

> Purchasing a certificate from a certificate authority adds another level of security and trust.

hahaha

well it certainly adds another layer of trust, I agree with that (you have to trust the CA not to mess up)

what's the story about DNSChanger?

xce

06:55

I wonder if any of the advertisers got in trouble for funding the crime..

the only way they could make money with DNSChanger is by people paying them to stick their ads in

what sort of contract would they be signing?

> at least US$14 million in profits

"If you are reading this page, it means you are NOT infected with DNS Changer" How does this webpage know?

seems like they have a list of infected IP addresses

> To stay safe from phishing and identity scams, download free SiteAdvisor

I just find that funny

http://cfdnscheck.cloudflare.com/

it just says "good"

not sure I trust cloudflare to accurately tell

xce

07:18

> Employers often monitor employee network access, even install keyloggers on their computers, etc.

is that really acceptable?

I mean the UK police can do it.. but employers?

Rory Alsop

08:27

@xce Yes, many employers have it specifically written into your contract that they can monitor anything they want, as the systems you are working on are owned by them. In practice they don't tend to monitor everything all the time, but if you are flagged for some reason or other (eg automated system tools identify that you are trying to access too many blocked websites) you may find that you are monitored more closely

xce

ah if it's in the contract that's not bad, you'd just know better than to do weird things like bank transfers during lunch breaks

Rory Alsop

oh - that you can definitely do, as the company does not have a right (in the UK anyway) to arbitrarily monitor your data. They could see your bank details, but if they used them in any way they would be in serious trouble.

you'd worry more about sending emails to terrorist groups :-)

xce

09:25

> Clam AntiVirus is an open source (GPL) anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail gateways.

wait how about just not running programs that people email you?

pretty good answer though raspberrypi.stackexchange.com/questions/1247/…

►

0

Q: How do I hide files inside of an image without using any stenganography tools?

I want to hide files inside of a picture. The files may be music or video. Is it technically possible to do this? If so, how? I searched on Google to find methods,suggest some methods for beginners ,i don't know where to start from,so guide me Important note : Modifying the size of an image up ...

steganography

a bit confusing.. is he actually asking "how do I do this without a tool that does it" or how do the tools that do this work?

Lucas Kauffman

10:20

@xce yea I tried giving him an answer, but I think is English is too poor to understand it

xce

also he seems to be asking "How do I do X without a way to do X"

Lucas Kauffman

@xce where are you from?

xce

UK

xce

11:09

-1

Q: How do I calculate CRC32 mathematically?

I want to calcuate the CRC32 (algorithm http://www.libpng.org/pub/png/spec/1.2/PNG-CRCAppendix.html ) using polynomials directly but I don't know how. I found the generating polynomial listed here https://en.wikipedia.org/wiki/Cyclic_redundancy_check (this corresponds to 0xedb88320 in the example...

crc

I don't know why I got downvoted !

xce

11:23

I really don't think anyone is going to answer my question but I can't figure it out myself too bad

Lucas Kauffman

11:46

@xce hackersdelight.org/crc.pdf

xce

13:33

I got a really nice answer but even if I follow it exactly I don't seem to get the right output s :(

Scott Pack

15:34

So the trick with finding collisions is to remember that you're not reversing the hash.

So the goal isn't to find x given MD5(x)

The goal is to find y such that MD5(y) ==MD5(x)

It just so happens that x may be y, but it doesn't have to be.

xce

yeah

Scott Pack

So by extension if we define x as MD5(x) then we have found the collision for MD5(MD5(x))

I leave developing the formal proof as an exercise for the reader.

xce

I don't understand

Lucas Kauffman

@ScottPack Do you know if in the formspring hack any of the passwords were resolved, it seems a bit exagerated that out of 420.000 leaked passwords (sha 256 with a salt) 200k were resolved

I read this is in an article posted by an OP, but I have my doubts

xce

with a salt? I'm surprised by that too then.. I'd like to know more

oh I get it now, given MD5(MD5(x)) you just need to find some y such that MD5(y) = MD5(x) then MD5(MD5(y)) = MD5(MD5(x)).. but you can't do that if you don't know x (or at least MD5(x)) (I've posed the question wrong by including x in the known information)

www2.mat.dtu.dk/people/S.Thomsen/wangmd5

this can find collisions but it says it's take 30 mins

but wikipedia says "This attack runs in a few seconds on a regular computer"

6

Q: Implementation of Tao Xie and Denguo Feng's MD5 attack

It is well known that MD5 is completely broken today - however, to understand the theory behind the attacks I am looking for an implementation of the collision attacks described in the 2009 paper A New Collision Differential For MD5 With Its Full Differential Path by Tao Xie. DengGuo Feng and Fa...

cryptanalysis implementation hash-collision md5

I finally understood what you said earlier about collisions Iszi, it only took me until now

Iszi

16:12

@xce There was actually a method for this posted on HowToGeek recently.

xce

the consensus seems to be that they didn't security.stackexchange.com/questions/17798/…

@RoryAlsop, don't know if it's exactly relevant to your question but I was in a computerology class where they had people make PHP files that sent GET data directly to SQL.. had my hand up for 20 mins but she refused to take let me proselytize about escaping quotes.

Don't know what they were trying to teach other than bad habits

Lucas Kauffman

16:27

@xce that's actually an important part

xce

what is?

I was just worried that if any of the students went home and created a website themselves it would be horribly vulnerable because they been shown the wrong way to do it

Lucas Kauffman

@xce using prepared statements

xce

I wonder why is the cert. here invalid forums.gentoo.org ?

Lucas Kauffman

it's valid for me

xce

ah maybe it was a CA I removed

oh no CAcert the free one, I wonder why mozilla doesn't include them

Iszi

16:45

Root cert isn't trusted by my browser either - default Firefox 14 config.

Chrome doesn't trust it either.

xce

I'm really not keen on the CA system, I need a firefox addon that tells me when a certificate changes

Iszi

...or IE.

xce

I don't care if it's still valid, if it's changed something's going on

Iszi

@xce Presumably, as long as it is signed by a trusted root cert which has not been compromised, the cert is really still good. They still need to be changed regularly because they expire. So, monitoring a cert for change alone will probably produce more false-positive alerts than it's worth.

Lucas Kauffman

it shows good for me on chrome (linux mint)

should I be worried :p

xce

16:58

I think it's fine

Scott Pack

19:41

@xce Glad to hear it. Hash collision issues are one of those things you just really need to take a step back and think about in a vacuum.

Sorry I disappeared like that. We've been at an ostensibly Celtic festival all weekend. Good music and reasonably good beer, but that was about it. I was desperately hoping to be able to show Madilyn a good ole fashion caber tossing.

Transcript for