Most everyone uses some kind of Multi-factor or '2 Factor Authentication". But our guest this week (who is going by "Matt" @infosec_meme)... Wanted to discuss some gotchas with regard to 2FA or MFA, the issues that come from over-reliance on 2FA, including some who believe it's the best thing ever, and we finally discuss other methods of 2FA that don't just require a PIN from a mobile device or token.
We also discuss it's use with concepts like "beyondCorp", which is google's concept of "Software Defined Perimeter" that we talked about a few weeks ago with @jasonGarbis (http://traffic.libsyn.com/brakeingsecurity/2017-011-Software_Defined_Perimeter.mp3)
This is a great discussion for people looking to implement 2FA at their organization, or need ammunition if your boss thinks that all security is solved by using Google Auth.
Youtube Channel: https://www.youtube.com/channel/UCZFjAqFb4A60M1TMa0t1KXw
Jay Beale’s Class “aikido on the command line: hardening and containment”
JULY 22-23 & JULY 24-25 AT BlackHat 2017
Join our #Slack Channel! Sign up at https://brakesec.signup.team
#Google Play Store: https://play.google.com/music/m/Ifp5boyverbo4yywxnbydtzljcy?t=Brakeing_Down_Security_podcast
#iHeartRadio App: https://www.iheart.com/show/263-Brakeing-Down-Securi/
Comments, Questions, Feedback: email@example.com
Support Brakeing Down Security Podcast on #Patreon: https://www.patreon.com/bds_podcast
#Twitter: @brakesec @boettcherpwned @bryanbrake
#Stitcher Network: http://www.stitcher.com/s?fid=80546&refid=stpr
#TuneIn Radio App: http://tunein.com/radio/Brakeing-Down-Security-Podcast-p801582/
What does MFA try to solve:
- Mitigate password reuse
- Cred theft - Someone stealing credentials from embarassingadultsite.com and turns they work out on a totallyserious.gov RDP server
- Phishing bad - same as above, except now you convince someone totallyseriousgov.com is legit and they give you credentials
- Getting to the point where old mate literally has more password dumps than time
- Honestly not going away, and combined with password reuse makes things pretty bad
- META: do we need to back this up with some stats? https://blog.barkly.com/phishing-statistics-2016
MFA / Bad things happening with that:
- AU Telecommunications provider sent multifactor SMS to wrong people
- RSA was owned years ago - and had to reissue a bunch of tokens
- On the plus side, obviously increased cost to attacker significantly to do that
- Phishing frameworks are everywhere
- Misc / Turns out U2F makes phishing kind of dead? (Read first amendment)
- Appears Backed up by the spec ( ‘Origin’ / https://fidoalliance.org/specs/fido-u2f-v1.1-id-20160915/fido-u2f-overview-v1.1-id-20160915.pdf)
- a) What does multifactor actually solve?
- b) Are we (infosec industry) issuing multifactor solutions to people just so people make money?
- c) Do these things give a *false* sense of security?
- d) What do you think about storing the token on the same box? Especially given an actor on the box is just going to steal creds as they’re entered.
Internal training / is this actually working?
Australia Post didn't think so
It's irritating and does break at times ( https://twitter.com/dguido/status/842448889697447938 )
C: I don’t like running some silly app on my phone
C: I also don’t like running around with a physical token
C: Embedding a Yubico nano in my usb slot leaves me with one usb port left
Also doesn’t solve when someone just steals that token
Does any of it matter:
Beyondcorp / "Lets make the machines state be part of the credential"
- Tl;dr of paper: TPMs, certificates and a lot of health checks - think of NAC on steroids
Is there some way we (not google) can make it so a credential is worthless?
Duo / “There's an app on my phone and it has context about what wants to do something right now”
Probably a step in the right direction
Kind of like some Aus banks which SMS you before transferring $X to Y account
Okta - (grab links to spec)
META // Does this actually solve it?
OAUTH - (grab links to spec)
Attacking OAUTH - https://dhavalkapil.com/blogs/Attacking-the-OAuth-Protocol/
META // It’s not MFA, but it makes the cost of unrelated compromise significantly lower
META // Engineering things to short lived secrets is a better idea
I think one of the better ideas being put out was by google in 2014, the ‘beyondcorp’ project (https://research.google.com/pubs/pub43231.html), simply put:
- The devices used everywhere are chromebooks run in standard mode rather than developer mode
- (Whitelisting For Free™)
- Everything is a web app
- Everything else can’t run due to app whitelisting built-in
- The device needs to also authenticate before the user can do anything, and is used as part of the judgement for access control engines
- Everything cares about the machine the user is using - It’s part of the credential
- Passwords are no longer important and it’s all single sign on
- Suddenly credential theft doesn’t matter
- The device uses certificates to attest to its current state, so stolen passwords without a valid device don’t matter
- As the device is a glorified web browser, and has app whitelisting, you’re not going to get code execution on it, malware no longer matters
- Caveat, someone will probably think of some cool technique and that’ll ruin everything
- See: Problem of induction / “Black swan event”
Obviously this is a massive undertaking and would require massive overhaul of everything, but it did look like Google were able to pull it off in the end. (https://research.google.com/pubs/pub44860.html).
Tavis is banging on LastPass again… https://www.ghacks.net/2017/03/21/full-last-pass-4-1-42-exploit-discovered/
Duo Security // Beyondcorp
More info on Beyondcorp
Misc// Hey google wrote a paper on U2F a while back
Touched on briefly / “Secure Boot Stack and Machine Identity” at Google - Servers which need to boot up into a given state (Sounds like U/EFI except ‘ Google-designed security chip’)
META // Patrick Gray (sic) interviewed Duo last week and talked about the same thing