The DMZ

9:29 AM

@MechMK1 I improved my answer, based on the RFC you cited and a comment under your answer. Thanks @ElieSaad for pointing me to this comment.

MechMK1

10:19 AM

@A.Hersean Now I don't feel guilty at all, since your answer is the better one

MechMK1

10:49 AM

I've been awake for a grand total of 49 minutes and I'm cursing more than a Tourette syndrome patient

A. Hersean

Thanks. This was a collaborative effort. I really like your reference to "cargo cult programming", which I hadn't learned about before in those terms.

MechMK1

@A.Hersean The first time I heard of Cargo Cult Programming was in reference to "Your password must not be longer than X characters"

As far as I recall, the reason for it was that some KFD had a maximum password length and would truncate the input after that length.

I think it was bcrypt, but I'm not sure

So someone said that the maximum password length was that length, and people just followed suit

The "real life" cargo cults are even more sad

Imagine being part of an indigenous tribe and spending weeks to build imitation airforce bases and parading them up and down, hoping that some "god" would send you an airdrop

A. Hersean

It was "crypt" in early UNIX and Linux.

MechMK1

Ah crypt it was

I wasn't that far off, just a semitone flat

Elie Saad

Yeah bcrypt implements crypt underneath which truncates on 72 chars (give or a take a char "\0")

MechMK1

10:55 AM

Didn't bcrypt have quite a lot of "revisions"?

A. Hersean

Ah... "\0"... such a jocker this one.

Elie Saad

Yeah, bcrypt is pretty secure, and the advise on the characters is just about truncation so as to not mislead people

MechMK1

2, 2a, 2x, 2y and 2b

Elie Saad

It still used the same function underneath

MechMK1

I feel like bcrypt 2x and bcrypt 2y could be Pokémon games

And yeah, bcrypt is nice, but I like Argon 2 better

Since it has freely definable memory costs

And memory hard functions are all the rage now

Elie Saad

10:57 AM

Yeah, you as a sec expert. As developers come to set that memory cost, they'd just fuck things up :p

A. Hersean

plaintext < base64 < crypt < bcrypt < scrypt < argon2id

MechMK1

That is true. Though there is a handy guide on how to do it correctly

more like plaintext < base64 <<<<< crypt <<< bcrypt < scrypt <argon2id

Elie Saad

I don't approve much about scrypt. It's either I go for bcrypt or argon2

MechMK1

Why not? I'm curious

Elie Saad

One sec

MechMK1

10:59 AM

5

Q: What are the minimum parameters for Argon2?

Argon2 is the winner of the Password Hashing competition, and currently recommended by OWASP for secure storage of passwords. One crucial step of Argon2 is determining the parameters used by the function. The current IETF draft titled "The memory-hard Argon2 password hash and proof-of-work function"

For the argon2 parameter thingy

Since I wondered the same

ccccccngfiduunuvkedcinutgegegvlrutfnueutfenn

Fucking hell

How often am I going to touch this shitty Yubikey today?!

Elie Saad

So about argon2, before you reach a certain memory limit, bcrypt is more secure. It's mentioned by hashcat creators

MechMK1

What is that memory limit? Just so I have some ballpark

Since I have no idea how much memory is "normal". 1MB? 16MB? 512MB? 2048 MB?

Elie Saad

About scrypt. Scrypt is not supported on all possible devices, and if it is, argon2 is. Scrypt did not reach its optimal use case for FPGA resistance. You'll need to tune it to do so, then again, just use argon2. Bcrypt is everywhere (since it uses crypt), is much easier to implement and fuck up.

Let me snatch that from a tweet, just need to find it.

MechMK1

Yeah, I agree about scrypt. But to be fair, if you already have scrypt implemented (before argon2 was a thing) then I don't see it as a high priority to switch to Argon2

I mean, you should do so eventually, and I suggest everyone to do a regular hash parameter update anyways

Like once a year or so

Elie Saad

Here you go

twitter.com/terahashcorp/status/1155129705034653698

And chickenman is part of the hashcat team as well. We worked hand in hand to build the password storage cheat sheet

> The Argon2 spec gives very detailed information.

Man this sentence gave me cancer. It is not about detailed information, it's about usable information. People fuck up cryptography because crypto people like to write in wizard-like grammar and functions

MechMK1

11:09 AM

So 256 MiB for memory, 4 or 8 for parallelization and iterations adjusted to match ~ >= 1 second execution time?

@ElieSaad That is undoubtedly true

Crypto people tend to write for other crypto people

And then I as infosec professional need to distill that down into usable information, and distill it even more for our clients

Elie Saad

mobile.twitter.com/jmgosney/status/1286683028895432704

Another one a bit related to when actually argon2 becomes more effective than bcrypt

MechMK1

Because yeah, I could tell some web application programmer "Just read the argon spec" and I might as well write it in chinese

Elie Saad

Hahaha true true

MechMK1

Actually had to do that yesterday, in a way

Elie Saad

Damn if you poked me I would have saved you a 100 reputation if I knew back in the days :p

MechMK1

11:13 AM

Who knows what nice things are coming in the future? Definitely nice to have more people with extensive infosec knowledge here. This chat got more on-topic in the last hour than it has been in the last few months

So thanks to @A.Hersean, I deprecated bettercrypto.org as our company's recommendation and instead we go for the SOGIS Applied Cryptography recommendations

Elie Saad

I used to lurk around in here, but I never saw people in it, so I just left. Now it seems more fun.

MechMK1

Though I have no idea why out of the 5 TLS ciphers in TLS 1.3, they only recommend three

Elie Saad

Yeah, that SOGIS recommendation was lit.

MechMK1

@ElieSaad I am trying to be entertaining, though it usually ends with us just messing about

Elie Saad

Not sure if you read the news on improving BGP security btw: wired.com/story/bgp-routing-manrs-google-fix

Sounds a bit like a save after so many fuck ups

MechMK1

11:16 AM

But why do they not recommend TLS_CHACHA20_POLY1305_SHA256 or TLS_AES_128_CCM_8_SHA256?

The only thing I know about BGP is that it used to be broken for so many years

Elie Saad

Huh, let me check that. That combination is secure AF (the first one). The second one ... eh, I'd go for GCM. CCM is based on another cipher and mashed up with a MAC

BGP is still broken

MechMK1

All 5 ciphers in TLS 1.3 should be secure

sogis.eu/documents/cc/crypto/…

Page 32

They only recommend AES-GCM and AES-CCM

Not that that would be a bad thing, mind you

But I liked that they included ChaCha20-Poly1305 so that we have at least one alternative should AES ever be broken (which is unlikely, but still)

Elie Saad

That's really weird. All of the ciphersuites provide FS, in which they should be included. You know what's the shittier part? TLS is the shittiest thing to look up information about, honestly speaking.

MechMK1

11:34 AM

It really is

I also have to admit I have no idea what an "8 octet ICV" in terms of CCM is

And why I would or wouldn't want to use one

And I keep touching my Yubikey oh my god

ccccccngfidubrgevlitgehcuitcggntgkfhvledhdfe

NO

FUCK

God damnit

Elie Saad

Hahaha how are you touching your yubikey so much. Or why, could be the right question

MechMK1

It's a YubiKey nano

So it's not the indented button design that requires conscious effort to press, but rather just a conductive surface on the outside

So whenever I touch the side of my laptop for one reason or another, it activates the YubiKey

For example, when I use the arrow keys, my pinky sticks off to the side, and just naturally touches it

And it's annoying as all hell

I'll quickly configure it to disable the short touch

Help my YubiKey Manager is drunk

Would you like to configure S̵̢͘h̨̛͡o͏̶̶̸r̶̵̸̢̨t̡̀̕͟͠ ̴͟Ţ̧͝o͏̡͢͏u̴̡c҉̴̕͢͝ḩ̕͢ or L̷̕͝o̴͜͏ņ͞͡g҉̢̛͞ ̧͘͢T҉̴o̷͘ų́c̨͞͝͏h̸̡͘͢͠ ?

The longer you look the worse it gets

If you want to run the Yubikey Personalization Tool on linux you have to build it yourself

MechMK1

12:12 PM

Finally

Elie Saad

12:28 PM

Hahah you kinda have to build everything on Linux.

Man I so want to be in that tool state. It seems to be the kinda mood for everyday issues

MechMK1

Building stuff is half the fun on Linux

Though dependency hell is real

Elie Saad

Arch packages are the real heroes for people that read the pkgbuild file

For those that don't they belong to the deepest level in hell

I'm not gonna lie I've spent my night hours for days just configuring distros which made me get to know them very well. Like one night I stayed up till 8am just because I wanted a certain config working, and you know how linux drivers work ... they don't

A. Hersean

On windows the Yubikey interface is also fucked up. It's compiled this way.

I though I was the only one with this issue, but it looks like it could warrant a bug report, because it's not listed on their bug tracker : github.com/Yubico/yubikey-manager-qt/issues

MechMK1

@A.Hersean It's actually a Virtualbox issue

@ElieSaad I wish I had the time to do that. I wish my day had 72 hours and I had no need for sleep

I already completely neglect social contacts aside from my family and my two friends

I swear, next time someone posts something off-topic and argues with "Well it's not working, and availability is part of the CIA triad", I'll post a question with "Why can I not find a girlfriend?" and then also argue that availability is part of the CIA triad and it should therefore be on-topic.

A. Hersean

12:59 PM

I use this application on windows 10 on bare metal, without any virtualization. The interface is less fucked up as in your screenshot, but it is still messed up.

MechMK1

@A.Hersean Weird, could be a bug then. On MacOS it works fine

And on linux too

A. Hersean

That's on my personal computer. I'll try it on my work laptop.

Elie Saad

@MechMK1 Oh, I no longer have that time. Long gone are those days.

Man that gf reference was a really edgy one, do you need to chat? :p

A. Hersean

It works fine on my work laptop.

MechMK1

@ElieSaad Nah, I'm fine :D

">tfw no gf" is a bit of a running gag here

@A.Hersean I got everything running on the VMs now. Too bad the new Yubikey Management tool doesn't offer all settings yet

Like for example, disabling sending an Enter keypress after entering a password

Elie Saad

1:14 PM

Btw, does anyone feel like contributing to WSTG? I am prepping a working group that is made up of practitioners and instructors to build a well refined version 5.
If anyone feels like adventuring with me/us, don't shy away. It could be helping set a structure, vision, being an author, reviewer, etc.

MechMK1

@ElieSaad I'd love to, but I don't know if I have the chops to do it

I've been in infosec since december 2018 and - I kid you not - I learned pentesting while doing a job interview for a position as pentester

Elie Saad

I've had a new web tester improve one test scenario and build another.

MechMK1

People say I was a bad student because I never learned for the test, but I learned how to bullshit my way through SO MANY THINGS.

Elie Saad

While some 20+ years testers just go and rant on twitter instead of helping projects grow

MechMK1

And I'll never forget the story where I corrected my professor during my bachelor's exam :D

That was so funny

Elie Saad

1:18 PM

We're listening, me and the ghosts in the room, go on with your story

MechMK1

So I had my bachelor's exam, and my teacher asked me how return values are passed around in C. My first answer was "Well, that depends on the calling convention". He said "Just assume cdecl" and I said "Then it depends on the size of the return value"

He said it was wrong, I said no, I am correct. He said "It is always passed through a register" - I said that I doubt a 1kb large struct will fit into a 64-bit register.

Awkward silence. "...that is true"

Elie Saad

A+

MechMK1

So yeah, either through a register, or if it does not fit, the calling frame allocates as much memory as required on the stack.

Though you really shouldn't be returning large structs directly

Imagine being 20 functions nested in and having to copy a huge struct around 20 times

github.com/Yubico/yubikey-manager-qt/issues/251

Had to add it

Elie Saad

1:57 PM

Huh, did you make that Enter key with markdown or did u use html?

MechMK1

<kbd>Enter</kbd>

That's supported on SE as well

Finds extensive use on SU as far as I know

Elie Saad

Yeah was just thinking if I missed on some markdown magic somewhere, "I was like no freaking way, no markdown supports this shit"

A. Hersean

This missing feature of Yubikey was very useful to decrypt containers. You could use the yubikey to append a long string after a weaker password and have 2FA for containers.

Elie Saad

Nice, so like a salt functionality using yubikeys. 2FA with a split-password kinda mentality

A. Hersean

TIL <kbd>

Elie Saad

2:15 PM

I went a bit more through the SOGIS document. I feel like they kept out those 2 protocols because they're not block ciphers, or particularly AES. Look at all cipher suites, only AES is mentioned for encryption.

I can't believe they just recommended PBKDF2, and didn't even set boundaries for the amount of iterations. Like no guys, it's not sufficient to say "Do enough iteration plz ktx".

MechMK1

2:31 PM

@ElieSaad Yeah, but PBKDF2 is cool because NIST and because it has incredibly good support

A. Hersean

At my work, we usually switch between the SOGIS doc and recommendations from my country's equivalent to NIST.

MechMK1

It depends here. The German BSI is a good source for recommendations though

A. Hersean

We can only use recommendations endorsed by our NIST equivalent (SOGIS being one). In part because half of the time we work for them.

Elie Saad

Because NIST is really non-biased :P

Yeah but PBKDF2 is even weaker than bcrypt against brute force attacks, and bcrypt has the same support

MechMK1

Not saying they're not biased or have "other priorities", but companies love those big names

Elie Saad

2:46 PM

Eh, a couple of years from now I am actually hoping that OWASP becomes something worth being in the fight of recommendations.

Oh yeah, definitely.

A. Hersean

The main advantage of PBKDF2 is its name: So, you want a "password-based key derivation function"? Let me think of one...

Elie Saad

Hahaha fabulous, fabulous ... starts sweating

MechMK1

3:02 PM

@ElieSaad >implying it isn't yet

Every other vulnerability I report in the course of a security assessment has a link to the OWASP Cheatsheet series

Because they're easy-to-understand for the target audience and get straight to the point

No mathematical proofs, no history of SQL injection throughout the ages, just "Do this, do that, enable this, disable that, use this library instead"

MechMK1

3:25 PM

Also by the way, did you know that Adolf Hitler won the election in Namibia?

ThoriumBR

5:22 PM

@MechMK1 I learned pagemaker 7 in 2003 while doing an interview to be hired as a main "graphics" man for the city hall. my background up to that was C/C++/PHP/Linux/Apache/DNS... I never even saw Pagemaker 7 before. I saw maybe pagemaker 3 or 4, years before, on a buddy's computer...

MechMK1

I assume pagemaker is some sort of "Make your own website, no code required!" thing?

ThoriumBR

I said "You use the version 7, I am more used to the older version, but I can manage..."

pagemaker is the tool to put text and images on newspapers

and make books and magazines

on my case, it was the official newspaper for the city hall, where they publish all laws, all official things...

MechMK1

6:54 PM

How would you format a single hard drive for a fileserver?

I planned on going ext4

Does anything speak against that?

Elie Saad

7:51 PM

Fresh out, WSTG version 4.2 is now out ladies and gents <3
https://owasp.org/2020/12/03/wstg-v42-released.html

ThoriumBR

8:49 PM

@MechMK1 I never had any issues with ext4, but I had with btrfs... so I always go ext4 when possible

this computer I am using is xfs, but so far, no issues..

MechMK1

@ElieSaad Nice!

Transcript for