Building a WebAuthn Click Farm(betterappsec.com)
betterappsec.com
Building a WebAuthn Click Farm
https://betterappsec.com/building-a-webauthn-click-farm-are-captchas-obsolete-bfab07bb798c
42 comments
It doesn't have to be 100% impenetrable to do its job most of the time and make it economically not viable with scale. Biggest issue is expecting a big chunk of users to actually have a FIDO2 token. They would have to be as ubiquitous as thumb drives or be built-in into phones and notebooks.
They are cheap enough to be ubiquitous (cheap ones can be had for $5 or so) and are already built into most phones, usually behind biometric verification (face or fingerprint recognition). Note the biometric only unlocks the FIDO2 private key in the phone’s TPM so biometric data never leaves the device.
Edit: just checked, a hyperfido mini sells for $9.99 CAD, so not quite five bucks but still pretty affordable.
Edit: just checked, a hyperfido mini sells for $9.99 CAD, so not quite five bucks but still pretty affordable.
> are already built into most phones
I don't know know about most unless you mean that most iPhones have it, but I wasn't aware it was built into mine, using Android 10. Just didn't put two and two together. My old smartphone on Android 9 didn't have it, I'm pretty sure.
I don't know know about most unless you mean that most iPhones have it, but I wasn't aware it was built into mine, using Android 10. Just didn't put two and two together. My old smartphone on Android 9 didn't have it, I'm pretty sure.
If this catches on I will make sure to buy and abuse a few of these tokens every month just to get them banned.
Doing a little trolling, 100k+ banned tokens at a time.
Doing a little trolling, 100k+ banned tokens at a time.
They don’t have to ban the tokens worldwide. They can start with your IP address.
The author points out that you can detect this by finding keys that have an unusually high signature counter. It becomes one signal amongst many, including IP addresses, user agents, etc that are correlated with bot-like activity. You still also get to track other things such as whether the user is moving their mouse and how realistic that looks, as well as past behaviour from that user.
I guess in the edge case where you are a legitimate but very heavy user of u2f keys and you happen to share one with a batch that is being abused for bad activity... you just get shown a regular captcha and continue with your day I guess.
The design of the system will surely reduce the percentage of people that get that experience however.
I guess in the edge case where you are a legitimate but very heavy user of u2f keys and you happen to share one with a batch that is being abused for bad activity... you just get shown a regular captcha and continue with your day I guess.
The design of the system will surely reduce the percentage of people that get that experience however.
>The author points out that you can detect this by finding keys that have an unusually high signature counter.
Ah that makes sense. No evil fun for me then :^(
Ah that makes sense. No evil fun for me then :^(
> Any opinions stated here are my own, not necessarily those of any past, present, or future employer.
Is it common to get in trouble with employers if this is not stated? I don't understand how a clearly non-work related blog post like this could land someone in hot waters with their employer. And even if it were I still don't understand how this statement alone can help.
Is it common to get in trouble with employers if this is not stated? I don't understand how a clearly non-work related blog post like this could land someone in hot waters with their employer. And even if it were I still don't understand how this statement alone can help.
The cost of including it is nothing, so on the chance it works, there’s only upside. If it doesn’t work, it’s the same as not having it but the cost to include it was nothing, so what reason is there to not include it?
I’ve seen it a fair amount in cyber security, especially if you are attacking or leveraging things that the company sells or promotes. Does it work? Anecdotally yes, but that’s not something I would bank on when releasing content that I feel needs a disclaimer.
Realistically, if someone wants to release something questionable, it’s preferable to work with the company prior to releasing it to find the verbiage they want used (similar to how many google GitHub repositories state they aren’t official Google products) and to understand if you will end up in hot water.
Realistically, if someone wants to release something questionable, it’s preferable to work with the company prior to releasing it to find the verbiage they want used (similar to how many google GitHub repositories state they aren’t official Google products) and to understand if you will end up in hot water.
Disclaimers have become magical incantations of a litigious, safety-obsessed culture.
They remind me of the safety labels on every day items containing often apparently ridiculously obvious warnings. Like electrical shock warnings on extension chords, or warnings to not put a ladder on a slippery surface.
For non US based individuals encountering these for the first time, they often were their first encounter with how different the US legal system really was/is from the rest of the world.
For non US based individuals encountering these for the first time, they often were their first encounter with how different the US legal system really was/is from the rest of the world.
Back, demon!
------------------------------------------------------------ WARNING: This product contains a chemical known to the State of California to cause cancer.
------------------------------------------------------------ WARNING: This product contains a chemical known to the State of California to cause cancer.
Is it consumer overprotection(in air quotes) culture at large or is it widespread runaway corporate politics
Plot twist: this post is a smokescreen for this person working remotely from home and needing to automate their U2F key pressing
Anything with a capacitive touch sensor can easily be automated.
How? I've never understood how to automate mouse presses or touch presses..
A simple mechanical touch device, with some kind of electromagnetic actuator.
I read about people who put a mobile phone in a fixed cradle, so that when the phone receives a call and vibrates, a band of steel attached to the cradle and with a bulb of conductive rubber on the other end locks into resonance and eventually hits a USB security token.
I read about people who put a mobile phone in a fixed cradle, so that when the phone receives a call and vibrates, a band of steel attached to the cradle and with a bulb of conductive rubber on the other end locks into resonance and eventually hits a USB security token.
For anti-bot / reputation systems, I think there's some value in the type of proof of space and time (PoST) algorithm used by the Chia blockchain.
Algorithms that do proof of idle resources switch the economies of scale so it's cheaper for the average user to participate in a system than it is for an industrial scale botnet or bad actor. IE: Scaling up costs real money while I have a ton of idle capacity sitting in front of me right now that I could use for proofs with an incremental cost of $0 to me.
Reputation and identity are ripe for disruption IMO and systems that prove idle capacity for a period of time are pretty interesting and in their infancy right now.
Algorithms that do proof of idle resources switch the economies of scale so it's cheaper for the average user to participate in a system than it is for an industrial scale botnet or bad actor. IE: Scaling up costs real money while I have a ton of idle capacity sitting in front of me right now that I could use for proofs with an incremental cost of $0 to me.
Reputation and identity are ripe for disruption IMO and systems that prove idle capacity for a period of time are pretty interesting and in their infancy right now.
You can actually buy FIDO keys wholesale for under $5 if the order is over 10 thousand units directly from the manufacturers.
"we would be able to quickly deauthorize manufacturers’ public keys at various levels of granularity."
The wording suggests they can't deauthorize an individual device's key, perhaps because they're not unique. So would they have to brick innocent devices that happen to be from the same batch as the attackers'?
The wording suggests they can't deauthorize an individual device's key, perhaps because they're not unique. So would they have to brick innocent devices that happen to be from the same batch as the attackers'?
They don't need to brick anything, they can just throttle those that match for a bit. Cloudflare are not in a business of blocking people, they just need to protect sites from being overwhelmed by bots.
>How I built a click farm to “bypass” Cloudflare’s CAPTCHA killer with some cheap USB security keys, an Arduino, and a bit of python.
I'm holding off buying WebAuthn stuff until fabled YubiKey Bio comes out since WebAuthn support is still so insignificant even compared to TOTP. But big players like Cloudflare supporting it might change the picture.
Apple is adding built in support to iOS: https://arstechnica.com/information-technology/2019/12/idevi...
It's already in there and ready to be used. I wrote a small IDP/authenticating proxy that uses it, and I can now log in to my stuff with FaceID. Or you can hold your old-school Yubikey up to the phone (that you enrolled on your workstation!) and use that instead. It is all very seamless.
There is absolutely no reason to have passwords anymore. You can't phish WebAuthn. It's built into every modern phone and PC. (There is some concern about it being easier to lose access to your account, but account recovery already exists for the three remaining people that don't use a password manager, and forget their passwords. If you're implementing it, do let users add multiple tokens, though. That way they can have a backup and easily enroll new devices.)
There is absolutely no reason to have passwords anymore. You can't phish WebAuthn. It's built into every modern phone and PC. (There is some concern about it being easier to lose access to your account, but account recovery already exists for the three remaining people that don't use a password manager, and forget their passwords. If you're implementing it, do let users add multiple tokens, though. That way they can have a backup and easily enroll new devices.)
This can be detected easily using
navigator.credentials.create.toString() != "function () { [native code] }"
Another way is to hook the Fetch API and log outgoing urls that are not authenticated by Cloudflare
Welcome to the start of another catch and mouse game.
navigator.credentials.create.toString() != "function () { [native code] }"
Another way is to hook the Fetch API and log outgoing urls that are not authenticated by Cloudflare
Welcome to the start of another catch and mouse game.
> This can be detected easily using
> navigator.credentials.create.toString() != "function () { [native code] }"
Are you sure? If I can replace navigator.credentials.create then surely I can replace its toString as well.
> navigator.credentials.create.toString() != "function () { [native code] }"
Are you sure? If I can replace navigator.credentials.create then surely I can replace its toString as well.
The author didn't do that in article. You sure can replace toString (which can be detected as-well).
> You sure can replace toString (which can be detected as-well)
Are you sure? You can make “native code” functions with .bind(null)
Are you sure? You can make “native code” functions with .bind(null)
(function(){ var a=function(){};a.prototype.toString=navigator.credentials.create.toString.bind(navigator.credentials.create); return new a();})().toString()Yes, I'm. For the record, your code can be detected easily using,
toString.name
Which results "bound toString" wheres real one results "toString"
toString.name
Which results "bound toString" wheres real one results "toString"
You can set toString.name
You can of course, but how about these?
(function(){ var a=function(){};a.prototype.toString=navigator.credentials.create.toString.bind(navigator.credentials.create); return new a();})().toString == Function.prototype.toString
yields false where
navigator.credentials.create.toString == Function.prototype.toString
yields true
(function(){ var a=function(){};a.prototype.toString=navigator.credentials.create.toString.bind(navigator.credentials.create); return new a();})().toString == Function.prototype.toString
yields false where
navigator.credentials.create.toString == Function.prototype.toString
yields true
Of course you can "just" modify Function.prototype.toString as well. And so on.
You know what that can be detected as-well. I'm trying to prove my point that it's endless cat and mouse game.
I think we're talking past each other then.
Informing people again, WebAuthn is authentication only, and does not substitute for signing, and client encryption.
It's been nearly 10 years in making, went through multiple complete spec rewrites, and endless erratas.
And for that bicycle reinvention attempt with extra feature of captcha, and DRM, Google killed the keygen tag in Chrome.
It's been nearly 10 years in making, went through multiple complete spec rewrites, and endless erratas.
And for that bicycle reinvention attempt with extra feature of captcha, and DRM, Google killed the keygen tag in Chrome.
Attackers are already willing to purchase a bunch of cell phones to emulate human behavior. 'Attestation of Personhood' with the use of a hardware key is completely broken.