back to article Want a better password? Pretend you eat kale. We won't tell anyone

People have a very poor grasp of what makes one password stronger than another, according to research conducted at Carnegie Mellon University (CMU) and published by the Association of Computing Machinery. The old rule that a password should contain letters, numbers and symbols mean respondents to the CMU's CyLabs study think …

  1. This post has been deleted by its author

  2. Anonymous Coward
    Anonymous Coward

    People's knowledge tends to be decades out of date, even techy people.

    Last time I checked 10^14 hashes per day was well within reach of one individual. This changes the rules completely. Rainbow tables go out the window. The state of the art is pattern analysis.

    Turns out people are so predictable when it comes to substitution, if told to add a number or punctuation they'll append it. If a capital letter they'll prepend it 95% of the time on a distribution distribution.

    You'd better hope your provider has stretched your password using PBKDF2/bcrypt, otherwise your efforts could be for naught.

    Personally I'd go for horsecorrectbatterystaple (diceware style picked cryptographically randomly from a world pool of 60,000) but with 5 words instead of 4. Either naked, or as the root of a Merkle tree, using either base64 or base85 compressed leaves as your actual passwords.

    Write your own password generator, then you can re-write it in any language with SSL bindings, from memory.

    1. Charles 9

      And what if you have a BAD memory? That's the biggest problem with passwords: we need to remember too many of them that you end up asking yourself, "Now was that 'correcthorsebatterystaple' or 'paperclipreactordonkeywrong'?"

    2. Seajay#

      Yep. That requirement of "Must include a capital letter and a number" is really annoying. It means that even my KeePass generated 15 character random strings are sometimes rejected because by chance they don't contain a number and as you say it's adding nothing to security.

      The new default password rules should be. You can't have anything out of the adobe top 100 passwords list (nor can you have the name of the site you're logging in to). If you want to enforce more security, then just keep increasing the minimum length.

      1. Robert Carnegie Silver badge

        A capital, a number, and no repeated symbol

        No repeats is another annoying condition that leads to reasonable pw choices being rejected.

        I therefore use N letters, the first being a capital (which may be too obvious or may not make much difference), excluding repeats, then two digits produced by looking at minutes and seconds on a digital watch which may or may not show the correct current time - whenever a new password is required.

        I may also add spaces in a regular pattern, just to help me read and type the thing.

        Now, how big should N be?

    3. Flocke Kroes Silver badge

      If only you could type commands and have each output to the next...

      strings /dev/urandom | less

      A pathetic 8 letter password picked from the command above gets you over 10^15 possibilities, and with a little practice you can memorise double that by typing it repeatedly, waiting and hour, typing it again, wait till bed time, type it again, then again in the morning.

      In real life, you will need at least a dozen passwords, and many of those will need their own date of birth, mother's maiden name and town of birth. So far, sites have been happy with answers like Miss Pertpwjb from Wdudlumy and echo $(($RANDOM/(32768/31)+1)) for a date. Put the sites' URLs, your user names, the passwords and all the lies in a file you encrypt with gpg, and remember to shred the file after you have proved you can decrypt the encrypted copy.

      The only difficult bit is mapping a new site's password validator. If the rules seem too cryptic, you can always try:

      Password'); DROP TABLE customers;--

      1. This post has been deleted by its author

      2. AndyS

        Re: If only you could type commands and have each output to the next...

        Good grief, some of the comments here sound like they've been lifted straight from the Autistic Nerd's Handbook. In tomorrow's episode of "how to solve real life problems in simple and accessible ways," we're going to address how to get rid of a spider on the ceiling using only 3 thermonuclear warheads, a submarine and the armed forces of 12 small nations.

        Write your own password generator? Practice typing passwords every hour? (for every site that needs one? I have over 50 in my password wallet since every e-commerce site needs one now, but many will only get used maybe once a year if even that). Get real.

        Reg's advice of using a password wallet is plenty. A random generator isn't bad advice, but you don't even really need one if you go with the correcthorse method, that way you can keep the app on your phone but easily type the password into a computer, for example.

        Remember, to get away from a bear, you don't have to run faster than a bear. Only faster than the other people running.

        1. Anonymous Coward
          Anonymous Coward

          Re: If only you could type commands and have each output to the next...

          What a shit analogy.

          Given the shear power of contemporary password cracking, replace your bear with the robot swarm from iRobot.

          You don't have to outrun the robot swarm, you just have to outrun the first 10,000,000 users.

          1. John Brown (no body) Silver badge

            Re: If only you could type commands and have each output to the next...

            "You don't have to outrun the robot swarm, you just have to outrun the first 10,000,000 users."

            UK population of ~60,000,000, the majority of whom are online and using passwords. Most are uneducated in security techniques. Keeping ahead of 10,000,000 is easy. Most "hackers" are after the low hanging fruit, obviously. But some are after the higher hanging fruit because it's likely to be juicier.

            1. Mark 85

              Re: If only you could type commands and have each output to the next...

              But some are after the higher hanging fruit because it's likely to be juicier.

              Or maybe the higher hanging fruit is less juicy and just more paranoid.

          2. Robert Helpmann??

            Re: If only you could type commands and have each output to the next...

            You don't have to outrun the robot swarm, you just have to outrun the first 10,000,000 users.

            Sooo... The swarm of robot bears is after the infinite number of juicy, low-hanging monkeys that are pounding their passwords on internet typewriters? Is there an equivalent to Godwin's Law for non-Nazi-inspired analogies?

            1. John Brown (no body) Silver badge

              Re: If only you could type commands and have each output to the next...

              "Is there an equivalent to Godwin's Law for non-Nazi-inspired analogies?"

              You've been Car-ed? Forded? Toyotad? (because car analogies abound and rarely work well)

  3. cmannett85

    Aren't strong passwords a little quaint nowadays considering the vast majority of hacks are from poorly patched systems or spear phishing?

    1. Mark 85

      ^^^ Upvotes... Why go after Joe when you can get all the Johns, Joes, Eds, going after a system.

  4. ops4096

    hmmm ...

    Here I was thinking that the algorithm published by The Intercept was sufficiently robust.

  5. Anonymous Coward
    Anonymous Coward

    What If It Wouldn't Be Pretending??

    I like kale, does that mean I shouldn't use it in my passwords? And what sort of recipe is password anyway? I'm guessing main course or possibly some sort of weird salad. Nobody ever puts kale in a salad.

    And why has The Register become a chefing forum? Are they taking over the BBC's recipes listing? I think they should.

  6. Anonymous Coward
    Anonymous Coward

    The same people believe L0vemetal is harder crack than Lovemetal, even though there's no real difference to an automated attacker.

    Yes there is. It's using numbers, so there's 62 options per character space, as opposed to (letters-only, capitals and lowercase, discounting punctuation) 52. So brute-forcing it is going to take that much longer. Doesn't make any difference to someone who's using a password list, if the password is on it (and that one looks obvious enough to be on a list)

    1. find users who cut cat tail

      It depends. As L0ve has smaller probability than Love, an optimal strategy will find it later. A real software may not, but I hazard to say most will. Which, by definition, makes it more difficult to crack.

      So, dunno why so many downvotes. They can be both cracked easily -- but that was not the question.

    2. Dave 126 Silver badge

      >So brute-forcing it is going to take that much longer.

      It really depends upon how the brute-force program is written. It isn't going to start at



      aaaaaac... ...zzzzzzz

      but will start with commonly used words and combinations - adam, angryant... ...zebra34.

      There is no reason the brute-force program isn't using common substations - l0ve, Lov£, l0v3 - if the author of the program has decided (based on analysis of leaked databases of real-world users' passwords) that such an approach will be faster.

    3. CaptainHook


      Yes there is. It's using numbers, so there's 62 options per character space, as opposed to letters-only, capitals and lowercase, discounting punctuation) 52.


      You are assuming that the attacker knows that your password only contains upper/lowercase letters, if he doesn't know that, he has to assume that the password contains other symbols. In that case and for a 8 character password, he is still looking at a 8^62 possible passwords even if your password is only using 52 possible symbols to brute force it. And even more possible symbols if the attacker has to assume that there are any of the printable ASCII characters, and even more than that if you start dealing with long lists of possible symbols.

      By having a password required to be X number of characters long as a minimum and must contain certain classes of symbols, you are actually weakening the strength of a _random_ password because you are telling an attacker a lot of about what the password can't be and thus reduce list of possible passwords. NOTE: The key word there is random, the problem is most people don't use random passwords.

  7. Filippo Silver badge

    I would love to just use a string of a half dozen random words or so, in the style of horsecorrectbatterystaple - except that 90% of sites demand non-alphas+mixedcase+numbers and frequently have a stupidly low length limit. Resulting in massive password reuse.

    1. AndyS

      Capitals on the words, replace all "e"s with "3"s and "a"s with "@"s Solved. Has been serving me well where strong passwords are required.

      1. Anonymous Coward
        Anonymous Coward

        NOPE. Modern password crackers do all those common substitutions, making your password a dictionary word, and therefore no stronger than the word it's derived from.

        1. Dave 126 Silver badge

          Substitution can have a place, as long as it not a common type. e > 3, 1 > l, o > 0 etc are no good.

          batteryhorse > cbuudszipstd

          'batteryhorse' is easy to remember, and 'advance each letter by one' is easy to remember. cbuudszipstd is slightly less vulnerable to a dictionary attack (though I note that zip is a word, and std is a standard (std!) abbreviation). If we started at batteryhorsestaplepurpleetc, there would be less chance of the output being mainly composed of dictionary words and fragments.

        2. AndyS

          I can see why I've been downvoted, so let me clarify. I know that password crackers know about substitution, but the problem is not the strength of the password (if I judge a 5 word, ~25 character password is strong enough for the application), the problem is the arbitrary restrictions put on passwords by the application. So if a long series of words is good enough, but it's being rejected, then do a simple substitution. Easy to remember, and achieves the purpose.

          My work, for example, has a ridiculous set of requirements for 2 of the 5 or so passwords I need for different intranet systems. All renew at different periods, all have different length restrictions, so they cannot be kept the same. Do I really care about keeping them secure? Not particularly - they're written down at my desk so that I can ring in and ask a workmate to log in if I need to. The restrictions are massively overbearing even putting aside the fact that they blatantly lower the system's security. In this situation, simple substitution is king.

      2. elDog

        O! BeJeezus - you h@v3 g1v3n Aweigh migh pr1vvy s3cr3t.

        Now, I'll need to update those little stickum notes.

  8. Ralph B

    Trust No One?

    > This is as good a time as any for The Register to suggest that the best thing to do is get a password wallet, and use a strong password generator rather than your own brain.

    Really? I've always thought these password wallets were rather like putting all your eggs in one basket. Can we ever be sure that these 3rd party solutions - whether open source or not - don't have any backdoors? Or be cracked? Or go bankrupt? Or otherwise disappear?

    1. Charles 9

      Re: Trust No One?

      If they're open source, though, they can be audited without need to consult the author. And if the author of an open-source project disappears, someone else can take it up (like with TrueCrypt to VeraCrypt). It's not like something that's been released can be UNreleased.

      1. Sandtitz Silver badge

        Re: Trust No One? @Charles 9

        "If they're open source, though, they can be audited without need to consult the author."

        They can, and that's the best part of open source.

        The worst part is that it's an academic point and just never done.

        A proper audit costs time and money and I'm not aware of open source audits except the Truecrypt case. Truecrypt is/was widely used, and while I'm not using it I'm still thankful for the effort. Unlike TC, OpenSSL is made by full time employees, funded by tech giants, used by countless companies to provide crypto and nobody cared about the quality until the horses bolted. Several times.

        All coders can read the source, but deciphering it and finding obvious or obscure vulnerabilities may be beyond them. If the code is implementing its own cryptography (yikes!) it would need someone really proficient in maths and crypto to spot failures or even to understand what's going on in the code.

    2. Seajay#

      Re: Trust No One?

      If you're really paranoid I can see why you might not trust a closed source, cloud-based, commercial service. But even then, if you pick a big one you know that the livelihoods of everyone who works there absolutely depend on them not fucking up the security so they're going to be doing everything they can to make sure it works and they almost certainly know more about security than you or I.

      The only way I can see an open source vault being compromised is if there is a keylogging trojan on your machine which manages to capture your master password and the database file. You were pretty stuffed in that scenario whatever password strategy you use but the vault has made it worse because even passwords you haven't used in the time before you notice the infection are captured.

      1. Dave 126 Silver badge

        Re: Trust No One?

        You could always use two password wallets, made by different teams, and combine their output:


        Of course this approach isn't as convenient as using just one password wallet.

      2. Anonymous Coward
        Anonymous Coward

        Re: Trust No One?

        Or they could, you know, put up something that LOOKS good but is garbage, wait for the money to come in, then make tracks to some non-extradition place before the crap hits the fan...

  9. Seajay#

    67 per cent figured a 50,000-guess-strong password was good enough

    And they're right, if it's a password for a system which doesn't contain anything of much value and will lock you out after 3 attempts.

    More seriously there is a huge difference between what makes a "good enough" password for accessing a remote system which is rate limited and monitored and what makes a "good enough" password for encrypting something which the attacker has access to.

    1. Adam 52 Silver badge

      Re: 67 per cent figured a 50,000-gruess-strong password was good enough

      Trouble is you don't know which attack you'll be up against - will your provider have their backend hacked or will it be external? If anything the former seems more likely these days.

      1. AndyS

        Re: 67 per cent figured a 50,000-gruess-strong password was good enough

        Well... you do, because you know how valuable the information protected is. New password for a shopping site you're probably not going to use very often? Low risk, low security, poor password is good enough. El reg forums? Reddit? Facebook? Twitter? Meh, nothing of real (monetary) value here.

        Paypal, amazon, bank or main email address? These can spend money directly and take over other accounts (in the case of email and password resets), so high security, decent passwords.

        Assume the worst (backend hacked or physical theft), then decide how much it matters. 90% of passwords don't need to be strong. My work password? Post-it under the keyboard in case I need someone else to log in while I'm away. Simple word, with a number I can increment every 3 months when required. This would be slammed by any "password guru" but in reality it's perfectly secure enough.

        1. Charles 9

          Re: 67 per cent figured a 50,000-gruess-strong password was good enough

          You forget social engineering and identity theft. They can use data from the less-valuable sites to make inroads into the more-valuable stuff. So since just about ANY site can be a stepping stone, you may have to assume your least valuable site is as important as your most valuable one (since breaking the former can lead to breaking the latter).

  10. M7S

    How crackable are alien languages?

    I expect that Ivan and Abdul have probably realised that some techs will try to hide behind Elvish and Klingon, and have tables to attack these at "state level", but for the rest of us is there any real risk in using these?

    Alternatively, how about "Bill & Ben"? Is "flobadobalob" too easy to crack?

    If so, am I better changing to "Soup Dragon"?

    1. This post has been deleted by its author

      1. Dave 126 Silver badge

        Re: How crackable are alien languages?

        You could try passwords derived from the Clangers, but I'm not sure how you would enter them.... Whaaawoooowah

        1. I ain't Spartacus Gold badge

          Re: How crackable are alien languages?

          No one's ever hacked my accounts:


          Password: weeewillmenditweeewillfixit

          1. Charles 9

            Re: How crackable are alien languages?

            Don't get too cocky. There may be someone out there who knows about Gabriel, Madeleine, and Professor Yaffle, too.

            PS. Dang. Talk about old memories. About 30 years by my recollection...

  11. Chris Miller
    Thumb Up


    For the Dean Martin reference in the subhead.

    1. Mark 85

      Re: +10

      Not why that got downvoted... maybe it was Jerry Lewis fan?

  12. Jtom


    I just use a standard template that's derived like this: start with the first or last name of a special person, a special date, and something specific about the website. Let's use Mary (mother's first name), 14 (married on Feb 14), and The Reg (this website). Then split the name and date into two parts, and structure the password like this: ma1TheReg4ry.

    Now use the same name and date for EVERY password, only changing the website info. So your brokerage account could be ma1Stocks4ry. It's quite easy to remember your passwords, and next to impossible for the typical jerk to break.

    1. Buzzword

      Re: Template

      This seems dangerous.

      Let's say I steal a password database from LinkedIn. I pick out all the passwords which contain the website name, e.g. yours is ma1LinkedIn4ry. Being a vaguely competent hacker, I'll go round all the banks trying ma1Barclays4ry, ma1HSBC4ry, ma1Paypal4ry, etc. Bingo: your money is mine.

      1. Francis Boyle

        Re: Template

        Whatever system you have it would be crazy to use for high value sites like banks.

  13. Just Enough

    fault of websites, not the users

    The fault here lies entirely at the implementers of crappy websites, with crappy password policies and crappy coding. They are the ones who have taught users that enforced policies like ...

    "At least 5 characters, but no more than 8. Two capitals, at least one number that's not at the start, a non-alphanumeric character that can't repeat anywhere, up to five other lowercase letters that can't be an anagram, and none of the following : @?#\./ or space."

    ... is actually how good passwords work.

  14. allthecoolshortnamesweretaken

    Record random drunken warbles in the wee small hours of a really good party. Transcribe phonetically. Pretty high entropy, probably not part of any existing rainbow table. Not that easy to memorize, though - but then again, you can't have it all. (Where'd you put it, anyway?)

    1. Dave 126 Silver badge

      arse... hic! feck... hic! DRINK! hic... girls... that would be an ecumenical matter!

      1. This post has been deleted by its author

    2. Charles 9

      Problem is we MUST have it all or it all comes crashing down. Problem is, you can't do the ARE, HAVE, KNOW business because many people only have the ability to produce an ARE (bad memory so they can't KNOW and lack of phones means nothing they HAVE).

  15. DropBear
    Black Helicopters

    Hmmm, proper passwords...

    Okay - what I'd REALLY LIKE TO KNOW is why any attempt to include a link to in this post results in either "bad gateway" from cloudflare or "invalid html" from El Reg, whereas even a single letter difference lets the preview perfectly through. Repeatably, any number of times. Should I start listening for these?!? -->

  16. DropBear

    ...and for the record, this is what I was trying to link: (yup, still doesn't let me link it properly)

  17. Anonymous Coward
    Anonymous Coward

    Passwords are irrelevant when resetting them is an insecure process

    If they look at all the passwords they consider "strong enough" they should then look at their security questions / answers. Anyone who answered truthfully might as well be using "password1". I'll bet only a few percent (across all users, it will be a majority of Reg readers) understand this and make up answers for those as well.

    I saw something highly disturbing recently from one of my credit cards. They are moving from the old system where you had a single security question and answer (but it was freeform, so you can put in whatever you want for both) to having five. That's more secure, right? Except that they present you with a menu of choices for the question - and a menu of choices for the answer! I've never seen anything so stupid before. Since they haven't "required" it yet I didn't do it, hoping someone will inform them of the error of their ways, but at some point they might require me to set those and lose my secure Q & A.

    I don't know exactly how their security will work, but if all it takes is answering one or two of those questions correctly pretty much anyone can reset my password. That's not a terrible thing since they'd have to get access to my email as well to make use of that, but it sure is sad that a freaking BANK would be dumb enough to do something like this!

    1. inmypjs Silver badge

      Re: Passwords are irrelevant when resetting them is an insecure process

      "understand this and make up answers for those as well"

      Being somewhat annoyed by the process and asked for a memorable word while registering for a site I chose


      Quite embarrassing/amusing when I later discovered it wasn't for online use but the answer to a security question asked on the telephone.

    2. Mark 85

      Re: Passwords are irrelevant when resetting them is an insecure process

      I've seen that on a couple of sites. What they appear to be doing is grabbing info from a credit report as they have answers for things like: "where do you live" that are places I've never lived but the ex-wife did. Yeah... she tagged my identity a couple of times and it took some real screaming at the credit report people to get it fixed. But they still show up from time to time.

      Color me pissed off... So pissed off in fact, I've gone back to those sites and demanded they cancel my account and remove all information on me or the lawyers will be knocking at their door. Get pissed enough and firm enough and they can and will remove you.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like