AOL's Embarassing Password Woes

An anonymous reader writes "AOL.com users may think they have up to sixteen characters to use as a password, but they'd be wrong, thanks to this security artifact detailed by The Washington Post's Security Fix blog: "Well, it turns out that when someone signs up for an AOL.com account, the user appears to be allowed to enter up to a 16-character password. AOL's system, however, doesn't read past the first eight characters." This means that a user who uses "password123" or any other obvious eight-character password with random numbers on the end is in effect using just that lame eight-character password."

Nothing new

[by Anonymous Coward]

It's nothing new, the BT Openworld webmail system had this unique bug/feature years ago. Wonder if they've fixed it....

Re:Nothing new

[sglider (648795)]

MySpace has that issue as well, past 10 characters. If you go to their signup screen, you can sign up with a longer password, but if you go to the secondary login screen, it will stop typing either after 10 or 12 characters.

Not alone

[bsane (148894)]

Solaris (up to Solaris8 anyway) has exactly the same problem, I wouldn't be surprised if its widespread on older systems.

One thing I find interesting though, way back before the internet was well known (1990 or so I think) and people paid for CompuServe or AOL or whatever, I had a CompuServe account and the original password was 'wrote*admiral' and it definatly required all letters to be correct

Re: same in the default install of solaris 10

[by Anonymous Coward]

Same problem in a default installation of Solaris-10 as well.

Re:Not alone, Apple too

[Branka96 (628759)]

Apple's OS X had the same problem until 10.3. See Apple KB article [apple.com]

Ditto NT4. Sort of.

[by Anonymous Coward]

NT4 broke a 16 character password and separately hashed the first and second parts so you could attack them separately. This is why passwords > 8 characters were recommended. Better than TFA, and (thankfully) fixed in NT5.

Worth remembering if you still have any NT4 servers in production.

Re:Ditto NT4. Sort of.

[kestasjk (933987)]

I think you've mixed something up.

The Lanmanager hashing system breaks the password up into two 7-char sized chunks, converts them to upper case, and hashes each separately, and XP still uses Lanmanager hashes if you don't explicitly tell it not to (by changing a registry setting).

The first 14 characters are still used in Lanmanager hashes though, so this is only a security hole if the attacker can access the hashes.

Re:

[Cygfrydd (957180)]

#PASS_MAX_LEN 8

Perhaps it's just me, but isn't that commented... meaning, the entire length of the password is hashed, and thus, significant?

Re:

[HBI (604924)]

You are right.

Re:Not alone

[TheRaven64 (641858)]

I don't know about Gentoo specifically, but on most *NIX systems the convention is to put the default values in the example config file, commented out. This shows the user what the defaults are, and shows that they don't need to be explicitly stated.

Re:

[Albanach (527650)]

Perhaps it's just me, but isn't that commented...

It's commented meaning the default applies. It also states the default is 8, so eight characters are significant.

Re:Not alone

[Cygfrydd (957180)]

# Ignored if MD5_CRYPT_ENAB set to "yes".
#
#PASS_MAX_LEN 8
...
MD5_CRYPT_ENAB yes

... which seems to indicate that the default behaviour is to ignore the password length cap altogether.

@yg

Re:

[ATMD (986401)]

I'm running an up-to-date Gentoo install, and have never knowingly touched that file. I just tried logging in as root, except typing only the first 8 characters of my password and then garbage. It didn't let me in.

Re:

[teh kurisu (701097)]

Forgive me if I'm being a spaz, but isn't that line commented out in your example? It also seems to be commented out on my Gentoo box, which leads me to believe that it's commented out by default as it's a file I've never touched.

Furthermore I tried su'ing on that machine with only the first eight characters of my root password, and was denied access. So I'm concluding that it's not a problem in Gentoo by default.

Re:Not alone

[PAjamian (679137)]

It's not just Solaris, here's part of /etc/login.defs on a Gentoo box:

# Number of significant characters in the password for crypt().
# Default is 8, don't change unless your crypt() is better.
# Ignored if MD5_CRYPT_ENAB set to "yes".
#
#PASS_MAX_LEN 8

# If set to "yes", new passwords will be encrypted using the MD5-based
# algorithm compatible with the one used by recent releases of FreeBSD.
# It supports passwords of unlimited length and longer salt strings.
# Set to "no" if you need to copy encrypted passwords to other systems
# which don't understand the new algorithm. Default is "no".
#
MD5_CRYPT_ENAB yes

Old DES crypt() hashing is only significant to 8 chars on any system. That's why modern systems (including Gentoo) use MD5 hashing by default which has no limit on the length of the password to hash. Notice that MD5_CRYPT_ENAB is set to "yes" above which causes it to ignore the PASS_MAX_LEN setting.

Standard crypt problem

[AEton (654737)]

This is not that unusual.

We switched to a new content management system and gleefully informed users that their new default password was (an organization-standard eight-character string) followed by their username.

We realized something was wrong when someone noticed that all the password hashes were the same.

(The fix: find a new better hash function.)

Re:

[Alioth (221270)]

Not only that, it either didn't have a salt or the salt was invariant.

Re:

[dohzer (867770)]

My recipe for hash definitely uses salt.

http://www.mspong.org/cyclopedia/cookery.html#hash ed_beef [mspong.org]

That's YOUR password?

[martyb (196687)]

"Me too!" :^)

Re:That's YOUR password?

[Jim Hall (2985)]

That's ok, I logged in and changed it for you. :-)

Spelling

[daybot (911557)]

No, whats really embarrassing is mis-spelling that very word in the title of a Slashdot article

Re:Spelling

[Hebbinator (1001954)]

Gotta get a spell check.

I spent all day yesterday giggling at "eLfavirenz" (its efavirenz- no L). While HIV/AIDS is far from a humorous disease, images of brazilian midgets with big ears and curl-toed shoes sneaking around with big bottles of pirated protease inhibitors kept jumping in my head.

For a second treat, google ELFavirenz and see the 260+ web sites that took the exact same text and put it up after /.'s error!

Ahh fixed the summary...

[The Living Fractal (162153)]

Well, it turns out that when someone signs up for an AOL.com account, the user has sold their digital soul to Satan.

I *still* cringe to this day when someone asks for computer help and it starts out with "Well, when I log on to my AOL..."

TLF

Even better

[AndrewM1 (648443)]

I can do this one better. I signed up for some game known as MapleStory a while back, submitting the password "DaedAEcarECel40s".

I quickly found that I could not log on to my account. I was wondering whether I misspelled my password or something, when I noticed (while reading the FAQ) in small print "Passwords must be 8 characters or less." Now, no warning of this was given anywhere on the sign up form.

In shock, I realized what the issue must have been. Sure enough, trying to log on with password "DaedAEca" worked like a charm.

Yes, not only did they not warn the user that there was a maximum on the password length while signing up, and not only did their form accept my 16-char password, but it actually would not let me log in with the full password. Man, I was pissed and confused for a while...

Re:

[by Anonymous Coward]

> It's the same thing with msn messenger. sign up with a really
> long password, and you're locked out.

But surely that's a good thing?

Radius?

[cluge (114877)]

I believe the original RFC for radius only looked at the first 8 characters. It would not surprise me if AOL was using a tried and proven radius solution, and never bothered to update. I'd be interested to know the results if one was to choose a long password and then

1. Log into AOL and only use the first 8 characters
2. Log into the AOL webmail and only use the first 8 characters.

This may indicate if the limitation is the sign in solution, or the entire userdb backend.

cluge

Re:

[juggler314 (556575)]

Man I noticed this years ago, wish I had thought it was important enough to write up about then maybe I could have had my own slashdot posting!

(and yes that...sickeningly...means I actually used AOL for some time...)

I had a problem logging in to the AOL webmail because it *does not* truncate to the first 8 characters and I *thought* my password was longer than 8. Thus logging into the AOL app worked fine, but I had to manually truncate to 8 characters to get webmail working.

I thought it was a problem on my

Its actually worse than that

[imunfair (877689)]

It's worse than they make out. Back in December 06 I posted a synopsis of how the password hashing on AIM works. They ALSO remove all the 'weird' (read: non-alphanumeric) characters. So your "eight characters" may actually be only six or four - since it cuts the password down to eight before it removes the weird ones.

They also don't hash passwords anymore in your registry from AIM6 onward. They encrypt them, but that's a lot easier to get around than hashing.

If you really want a more detailed explanation you can take a look at the 12/29/06 and 12/30/06 posts on this page - http://tsourceweb.com/ [tsourceweb.com] - but what I already mentioned is the crux of the issue. (We all know people on Slashdot dont like to read articles anyway ;)

Worse than it sounds?

[Jugalator (259273)]

For random passwords, I guess 8 characters are still OK, but it's worse if you pick "smart" combinations of words and numbers, like "computers4life" or "jennifer2007". With dictionary attacks adapted for these lengths, they'd only need to check for the first 8 and it would be "computer" and "jennifer" in this case. If you further adapt the attack to only look for e.g. ratios of 4:4 with first 4 being a word and remaining 4 being random, and so on for 5:3, 6:2, 7:1, and 8:0, you also catch circumstances where users have picked passwords like "love4u2007", which would be caught in the "4:4" attack as "love" + "4u20". Maybe that's still secure enough, but this sounds a bit risky when using word passwords, even when mixing with numbers to avoid dictionary attacks, especially with this limitation.

This is AOL we're talkikng about...

[ZeldorBlat (107799)]

Do you really think the type of people who use AOL would use a password longer than eight characters anyway?

At a certain university,

[MulluskO (305219)]

At a certain university, this was also the case.

The flaw in question seemed to apply only to a web mail client which they are in the process of phasing out in favor of an open source solution, which is pretty interesting because it's the first I've seen which has support for S/MIME.

Presumably, the older system will be brought off line soon, as the flaw has been known for some time.
When signing on in front of people who didn't know about the flaw, it was fun to make them think you had a password in excess of

AIX

[Sp00nMan (199816)]

The latest AIX 5.3 has this same stupid limitation too. It's driving us nuts at work cause we authenticate to Active Directory which supports long passwords, but AIX only cares about the first 8. Ridiculous.. We had to purchase SpecOps and force AD to limit to max of 8 so that users would be forced to have a unique password everytime. We contacted IBM and they said they had no plans on fixing this.

Found this last year.

[BrianRagle (1016523)]

I believe I encountered this last year when I was trying to set my wife's AIM account up on her iChat client. She has been typing the long version of her pass into the AIM client, which apparently wasn't reading past those first 8 characters. When we tried it in the iChat client, it kept spitting it back out as being incorrect. We eventually had to change her pass to a shorter one to get it to work.

Here's Why

[N8F8 (4562)]

AOL management must make the same assumptions about AOL hackers that the rest of us do about AOL users.

Mitch Hedberg

[Himring (646324)]

Reminds me of that Mitch Hedberg joke:

"You know when a company wants to use letters in their phone number, but often they'll use too many letters? 'Call 1-800-I-Really-Enjoy-Brand-New-Carpeting.' Too many letters, man, must I dial them all? 'Hello? Hold on, man, I'm only on "Enjoy." How did you know I was calling? You're good, I can see why they hired you!'"

RIP Mitch

Flat Out Wrong - Read

[madsheep (984404)]

First, this article is flat out wrong and I challenge you to try it yourself. The AOL service will only allow up to 8 character passwords for e-mail related items. My password for my AIM clients has always been greater than 8 characters and I *cannot* log into anything without typing the entire password. This includes any web-based service at *.aol.com (primarily controlled by my.screenname.aol.com). I am a bit perplexed at where this article is getting its information.

br/>
A few test cases to pay attention to:

1) Sign up for an AOL mail account https://new.aol.com/freeaolweb/?promocode=814322&n cid=AOLAOF00020000000602 [aol.com]

Notice it only allows you to choose a password that's 6-8 characters, just like the AOL service itself. So now try and login with your password that's 6-8 characters, but add a few more. It lets you in right? Ok, so do this... reset/change your password now. Click "Forgot my Password" or whatever the link is called. Go through the questions and set a new password. Oh wait, notice it only lets you pick a 6-8 character password.

What does this mean? It means for AOL-service based/AOL-mail based accounts, they only allow 6-8 characters for the password! Who cares if it accepts extra characters. There is a 6-8 character limitation. It's absolutely irrelevant that it accepts additional characters.

They seem to be confusing this with AIM-only based accounts, which allow up to 16 character passwords and DO NOT allow anything more or anything less than the *EXACT* password. Try it yourself. If my AIM password is "pCv921!$z" it will reject me if I put "pCv921!$" and it will reject me if I put "pCv921!$z44". This is not that big of a deal and certainly isn't embarrassing. This is flat out a difference in AOL's mail-based system vs. AOL's AIM-based system.

Want to know a big shocker about AOL's mail-based system that they didn't figure out and report on that *is* embarassing?

These AOL.com (mail-based) and AOL-service based account are *NOT* case sensitive. That's right, try and make your password with some uppercase letters. It doesn't make a difference if your 6-8 character password has uppercase letters or not. It doesn't recognize it! I didn't check but I don't believe it recognizes special characters either. So your character set is a-z0-9.

Chew on that. Steven :)

Embarrassing?!

[morari (1080535)]

What exactly about AOL isn't embarrassing?

Old adventure games

[Michael Woodhams (112247)]

Old text adventure games were often like this. You'd type in an entire sentence, but the computer would only look at the first three letters of the first two words. I remember using "drink white paint" to drink the whiskey. (This was back when the final resting place of outdated computer games was not the $10 bargain bin, but rather having the entire source printed in a computer games magazine so people could type it into their Apple II.)

I think that Infocom, being the class act of text adventures, didn't suffer this "feature".

Re:No way.

[creimer (824291)]

Nope. At some companies I worked for, the most common passwords are "password", "hockey" (I have no idea why), and "yousuck" (Windows machines). The opposite extreme is companies with password Nazis who insist that your password be a certain length, follows a certain pattern (capital letters, lowercase letters, numbers and symbols) and minimum length (eight or more characters), must be changed every 90 days, and you can't reuse the last 500 variations of the same password based on your name.

Re:No way.

[Bastard of Subhumani (827601)]

... thus pretty much ensuring that you write it down.

Re:No way.

[General Wesc (59919)]

I used to tell people not to write down their passwords, but after dealing with people losing their passwords all the time, I changed my tune. I think this makes a good point [berylliumsphere.com]. There are some passwords I won't write down, but if I can carry hundreds of dollars, keys to my house and car, and credit cards with over a total credit line over 10 000USD in my pocket.

Preferably, one would just write down a hint, of course. And not on a sticky-note on the monitor.

Re:

[thogard (43403)]

It changes authentication from something you know to something you have.

Re:

[Rakishi (759894)]

Under the keyboard? That's a rarity, mostly they seem to be stuck to the monitor.

Re:

[cp.tar (871488)]

Now those are people who do not understand the way people think. Mathematicians, not psychologists.

And they are the reason social engineering works so well.

People like having one, maybe two or three passwords.
So instead of making them change passwords regularly (and do note the analogy of having to change your front door lock every two months!), make them create one relatively secure password and drill them to memorize it, never, ever reveal it to anyone and never ever write it down.

Changing passwords d

Re:

[myowntrueself (607117)]

The opposite extreme is companies with password Nazis who insist that your password be a certain length, follows a certain pattern

I've seen ones where they specify things like 'must be 10 characters long, contain 2 symbols, 2 numeric characters, 2 uppercase'. They don't seem to realise that they are actually *reducing* the complexity of possible passwords.

If a cracker knows that a password *will* contain, eg, 2 non-alphanumeric characters plus 2 numerals plus 2 upper case characters and the required length

Re:

[Tim C (15259)]

Given that I saw exactly this behaviour on a Solaris 8 install at work a few months ago, no, I completely believe it.

Of course, *then* I was shocked...

Re:

[by Anonymous Coward]

> So that's the same as in most (all?) Linux distributions by default.

Was that a question or a statement?

No linux distro that I have used in the past 8 years hashes only the leading 8 chars of a pass phrase. Even so a strong 8 char password is still a strong password (eg: *_Jilt3d) or even better with non-printable chars.

Re:

[Bastard of Subhumani (827601)]

Even so a strong 8 char password is still a strong password (eg: *_Jilt3d)

It isn't if you're relying on the part after the eighth character to make it strong and the system is silently ignoring that part.

Re:

[ettlz (639203)]

still a strong password (eg: *_Jilt3d)

Trying to tell us something?

Re:

[julesh (229690)]

So that's the same as in most (all?) Linux distributions by default.

Not since some time around 2000 when all of the major distributions switched from DES to MD5 authentication. Some major Unix vendors do still have the issue, though.