Slashdot Mirror
Ashley Madison's Passwords Cracked, Soon To Be Released
New submitter JustAnotherOldGuy writes with some news that might worry anyone caught up in the Ashley Madison data breach. ("Uh-oh," he says.) Now, besides any other possible repercussions of having one's name on the list of account holders, there's a new wrinkle. The passwords used to secure those accounts were theoretically robustly protected with bcrypt. However, as Ars Technica reports,
That assurance was shattered with the discovery of the programming error disclosed by a group calling itself CynoSure Prime. Members have already exploited the weakness to crack more than 11 million Ashley Madison user passwords, and they hope to tackle another four million in the next week or two.
This would matter much less if passwords weren't so frequently re-used.
Damn! They cracked my password already.
Have gnu, will travel.
Instead of cracking the slow bcrypt hashes directly, which is the hot topic at the moment, we took a more efficient approach and simply attacked the md5(lc($username).”::”.lc($pass)) and md5(lc($username).”::”.lc($pass).”:”.lc($email).”:73@^bhhs&#@&^@8@*$”) tokens instead. Having cracked the token, we simply then had to case correct it against its bcrypt counterpart.
saved tokens for automatic login, in MD5
Be or ben't
This kind of stuff is the reason I never re-use passwords across services. All my passwords are randomly generated and stored by KeePass. Sure, it's a little less convenient to have to unlock the password safe in order to get into services, rather than just type in something you've already memorized. But, it's the only way to be sure that having your password compromised on one service won't compromise an account on another service. Even if the service isn't externally compromised, there's probably a lot of systems out there where employees (DB administrators, programmers) can gain access to the passwords from various methods such as logs or unaudited code.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
And that's when anonymized data is no longer anonymized.
We only publish anonymized data......but you can query down to all white men, aged 24, born in Wisconsin, living in New York city, own an Apple MacBook Air, earn $60k/yr, graduated from NYU, has a degree in Marketing, etc.
If you can add enough data points, your set gets down to one person -- even though that data is anonymized.
Having the salt available would still mean that all the hashes would need to be reversed with rainbow tables or brute-forced. These being bcrypt hashes with a built-in salt, having any "extra" salt available would hardly help.
Using the same salt for every hash is far less secure than having one for each hash too.
"When information is power, privacy is freedom" - Jah-Wren Ryel
The Ashley Madison system stored an MD5 hash of the lower-cased username and password on the user's computer, so that they could revisit the site without having to reenter their login info.
Computing MD5 hash values is much faster than computing bcrypt() values, the hackers already had the username, and both fields were lower-cased.
They just brute forced the MD5 hash until they got a match. About 90% of the MD5 passwords matched exactly (ie - the passwords were already in lower case), of the remaining 10% they tried uppercasing the individual letters of the password until it matched.
Security is hard. Basing the MD5 hash on a reduced-space plaintext password was the fundamental error.
Also there were some administrative lapses. They changed password hash algorithms, and then forced users to change passwords at next login. Many users hadn't logged in in several years, so this left a lot of old, insecurely hashed passwords around.
Generally poor security for such a sensitive site. Makes me wonder how good other popular sites are at security.
We really should figure out this security thing.
Perhaps an open-source fixed-function password keeper (as Mooltipass) in separate trustable hardware would work?
The Ashley Madison developers did a lot of things right. They even used strong encryption for the passwords. They improved their security over time. Yet, a couple of security bugs ended up taking the company down completely. With security, if you score 98 and the attackers score 2, finding two vulnerabilities, the bad guys win. Bugs happen. Security bugs are not okay, however.
I have a lot respect for good application developers. The blend of skills required is fairly comprehensive - UI design, database, understanding scalability, etc. With your wide breadth of skills, are you fine folks starting to understand that security is HARD, and requires a depth of understanding? That it's one of those things where it is wise to get expert assistance?
I've been programming professionally for 20 years, and I'm pretty competent; yet I'd never design and implement my own filesystem, because filesystems are HARD to do well. There are maybe a dozen people in the world who have the specialized knowledge and experience needed to design and implement a filesystem that rivals btrfs or even ext4. I KNOW that I don't have that specialized skill. One of my best friends has also been a professional developer for 20 years. Every month, he asks me about a security related issue, because he knows that he's not a security specialist, and that bugs happen, but security bugs are not okay. Will you let those of us who live and breath security 24/7 lend a hand before you release it next time?
$2a$12$p9Ctp8EvU1x9jc09dqslHeGxS/Ytu464Xs5Yn1/AkqMSqAAN.4coa
The salt is p9Ctp8EvU1x9jc09dqslHe, the 22 characters that follow the $2a$12$. If you want to crack this password, make a guess, use bcrypt to combine it with that salt, and if they match you've cracked this password. This one is not hard to guess.
It still gets me.
You run a huge operation, with thousands of users and millions of dollars flowing through it.
At which point do you need to stop and think "Actually, I need a server that does NOTHING but authentication, isolated from everything else?"
Literally a machine that can only communicate Yes or No and maybe a tiny token and every communication to it can only be replied to by yes, no, or issue of a temporary token (which can only be verified by the same machine answering yes or no).
Changing passwords is a rare, deliberate, easy-to-audit and unusual act - you could literally have a guy who has to press a button to okay each such action. Apart from that, an application has absolutely no need to do anything more than pass on info to a server that can reply yes or no. Whether that's from a initial password login, or checking a temporary token issued, that's all it needs to do.
It's not the be-all-and-end-all - you can compromise the interface and wait for a user to log on and thus capture a successful transaction - but this outright theft of every login detail and a list of things that, given time, can be turned back into passwords shouldn't be happening, should it?
I mean, quite literally, a serial cable should be able to handle such information on the scale of a half-decent sized website. Is this user 1's password? No. This is what user 2 claims his password is, can I get a token for that valid for the next hour? Is this token valid for user 2? What more beyond that do you need to program against to authenticate absolutely anything imaginable?
And even password updates - they operate on the same principle as the way that admins cannot see their user's passwords. We can update them, but we can't actually see what they were and the very act of updating them locks out (and therefore alerts) the genuine user.
Isolate this stuff. Seriously. An entire network that is air-gapped from your real network and literally the applications either side can ONLY communicate over a protocol that contains the bare minimum of commands. You could do it with an embedded device. Why are places with millions of dollars of business storing anything on a device that can be read back en-masse by even their own staff, let alone a compromised machine on the company's office network or similar?
Ya, like sirber said above. In another story, they say:
The source code led to an astounding discovery: included in the same database of formidable bcrypt hashes was a subset of 15.26 million passwords obscured using MD5
and...
"Instead of cracking the slow bcrypt$12$ hashes which is the hot topic at the moment, we took a more efficient approach and simply attacked the MD5 ... tokens instead."
I thought I had seen a story about a problem with PHP's bcrypt implementaion not too long ago, but I can't find anything on it now so I might have misread something.
Nevermind that, she's missing her period.
If you think I voted for Trump because of this post, you're wrong. I voted for Dr. Jill Stein of the Green Party. Again.
But that was not the problem.
They converted username & password to lowercase, and stored username, MD5("username::password") additionally to bcrypt2("username::password"). The MD5 hashes were resolved now, which is what this article is about. If they had not unnecessarily stored the MD5 hashes (probably a legacy field in the database, because only present for 11 of 36 million users), there would be no problem. Converting the password to lowercase was also unnecessary. The bcrypt2 passwords remain uncracked -- the remaining 25 million user entries still are secure as far as we can tell.
Article here: http://arstechnica.com/securit...
NB: The message above might reflect my opinion right now, but not necessarily tomorrow or next year.