Bitcointalk Forum Hacked | Hacker News Reader

Bitcointalk Forum Hacked

pastebin.com

53 pointsrdl26 comments

It should be noted that it appears to be a persistent hack that has been around for some time. The defacement was quite a lengthy animated video. The attacker had at least access as the PHP user, which of course has database access.

There's a video of the defacement here — http://youtu.be/LKrOHAfMdxI

A 50BTC bounty has been offered by Theymos (author of the advisory), awarded to anyone who can find the source of the compromise. It's near impossible for an outsider though as despite it running SMF, there's been extensive modifications which may have been the source of the vulnerability.

Source on the bounty: http://www.reddit.com/r/Bitcoin/comments/1nmdq4/bitcointalk_... also with an update:

    Update: It's unfortunately worse than I thought. There's a good chance 
    that the attacker(s) could have executed arbitrary PHP code and therefore 
    could have accessed the database, but I'm not  sure yet how difficult 
    this would be. I'm sending out a mass mailing to all Forum users about this.

Personally I suspect that it's due to misconfigured nginx which allows uploaded files to be executed, a user in #bitcoin demonstrated they could upload a PHP file using an exploit in the avatar uploading system, although theymos did have someone looking at the nginx configurations about 6 hours ago and hasn't mentioned finding anything.

Looks like end-point security, the (usually PHP) applications the server is running at, is a weak point in the Tor ecosystem.

Generally, I would rant on PHP for this, but seriously now, had they been on jvm, a complicated application like a forum talk with avatars and all other fancy features, would probably still have holes in it.

If you're running a Bitcoin site and you're using off-the-shelf PHP, you're probably going to have a bad day.

The gov't sure has been busy considering it's supposed to be shutdown.

/puts on tin foil hat

There are approximately fifty zillion "what exactly gets cut off in a government shutdown" articles floating around, meaning it's not terribly hard to actually find out what's still operating and what isn't. Plus, anything that was going on before the shutdown doesn't get magically back-in-time-un-happened by it.

Because I was being 100% serious.

http://pastebin.com/4Hf37D9H

:) nice to have the one responsible confirm that the db was dumped.

7500 rounds of sha256 :)

Anyone know how long it would take to crack a 8 character password on a laptop.

Yes, use a hash algorithm which the whole bitcoin community races to come up with ever faster ASICs. Brilliant. scrypt/bcrypt anyone?

Hardware ASICs are specialised for bitcoin. They wouldn't be useful for hash cracking unless they were modified.

It isn't using sha256, it's using sha256CRYPT.

It's not the same thing, it's designed to be used for password hashing

it's 7500 rounds of sha256crypt

I just ran it on a macbook pro with 8gb ram and 2.5ghz i5 and it took 0.014s per hash. It would take about 40 years to brute force 50% of the all lower case 8 letter passwords

(The same computer doing the same thing with sha256 would take 2.6 days)

http://hashcat.net/oclhashcat-plus/ gives 1003 Mc/s for one AMD hd7970.

Divided by 7500 that's 134 kc/s. 18 days to search through all alphabetical 8 character passwords, and 244 to search through all alphanumeric 8 character passwords.

Realistically an attacker would have several cards and use rule based dictionary attacks. Using a secure password hashing algorithm will not help you against weak passwords.

Not a good week for Bitcoin.

last time I checked, the bitcoin protocol was doing exactly as it was supposed to.

This sort of reply is a classic sort of reply in which a hacker ignores the social context of technology.

Bitcointalk.org users data for sale?

see http://pastebin.com/6S2H21eF

No one has purchased this (yet).

https://blockchain.info/address/1t4k5Y4NFFkvEAqBPymW2PDA5pZq...

Probably a scam.

why is 7500 rounds the optimal number?

Bitcointalk is a busy site. It's could be the most they could get away with without burdening their server

They you just add a CUDA enabled vga card to the server and all of a sudden, you can handle a lot more load.

Sure, and that's one of the problems with server-side key strengthening. You start off making a cracker do a thousand or million fold more work. They go and get better technology, like a GPU or an ASIC. So then you have to go upgrade your own hardware, or skimp and make the user wait for 1, 10 or 30 seconds while they login and you strengthen their password. It's an arms race and not something a star-topology (single server serving potentially thousands of users) is going to scale well on. You want those cycles rendering your HTML or doing something valuable.

Client-side password stretching and memoization simply makes more sense.

It's not, PKCS5 v2 just makes a statement about doing as many as is tolerable to you, with 1000 being a minimum. http://tools.ietf.org/html/rfc2898#section-3

IIRC, some hashing algos lose strength when this is done, but I can't remember what one.

Also, I am not a Cryptographer.

Reply on news.ycombinator.com