passwords strong enough for bruteforce

[her34]

edit: thanks for responses. i'd generally consider 100 years impractical



i read somewhere how the ever increasing power of computers is causing the length of a "safe" password to increase beyond what typical people would want to deal with (less than 10 characters)

using uppercase, lowercase, and numbers, how long would the password have to be to make bruteforce impractical by anyone (person, organization, etc)?

or does anyone know how many passwords/second the world's fastest supercomputer would be able to calculate?

 

[blackllotus]

The supercomputer entry on wikipedia suggests that no supercomputer has achieved computing speeds greater than 1 petaflop (no general purpose one has achieved above 360 teraflops). I think it is to safe to say that testing a password string takes longer than a floating point operation so we can use 2^50 password tests per second as an upper bound.

Now we can calculate how long a password string must be for it to take a year for supercomputer that can test 2^50 passwords per second to crack it. This hypothetical supercomputer can test 2^50 * 60 * 60 * 24 * 365 passwords per year, so with 62 different characters (26 lowercase + 26 uppercase + 10 digits) you should need about log(2^50*60*60*24*365)/log(62) = 12.7 -> 13 characters.

EDIT: Guess you don't need this anymore...

 

[JustAnAverageGuy]

i read somewhere how the ever increasing power of computers is causing the length of a "safe" password to increase beyond what typical people would want to deal with (less than 10 characters)

using uppercase, lowercase, and numbers, how long would the password have to be to make bruteforce impractical by anyone (person, organization, etc)?

or does anyone know how many passwords/second the world's fastest supercomputer would be able to calculate?

Just remember that every additional letter increases the difficulty at an exponential rate.

Using your given choices (26 upper + 26 lower + 10 numeric = 62)

Worst case scenario for brute force password formula:

Time = numberOfChoices ^ lengthOfPassword

a, b, c, ..., z

aa, ab, ac, ..., az, ba, ... Aa, Ab, ... , ZY, ZZ
etc...

62 ^ 0 = no password, one try
62 ^ 1 = one character, 62 tries
62 ^ 2 = two character, 3844 tries
62 ^ 3 = three character, 238328 tries
62 ^ 4 = four characters, 14776336 tries
62 ^ 5 = five characters, 916132832 tries
and so on and so fourth.

You'll have to define impractical. A few minutes? an hour? Five years? Any decent security system will also have some sort of brute force lock out. (i.e. 3 tries every 90 seconds)

You might say that 128-bit (read: letter) passwords are safe. By the time anybody guesses it correctly, the information contained within will be obsolete.

If I'm not mistaken, if you put a 256 character password in a brute force attack, it would take all the computers in the world longer than the time the universe has been in existance to crack via brute force

- JaAG.

 

[CycloWizard]

This isn't necessarily related, but I had a question somewhat along the same line: what difference (if any) would using a quantum computer have on the time required to find the password? I know very, very little about the idea of quantum computing, but would it be right to say that it can test multiple states at a time (given the right software)?

 

[CSMR]

But remember the fastest supercomuter will be faster in 100 years' time. You could assume moore's law still holds and do a little integral. (But of course this law could be unsustainable or alternatively quantum computers could come out and be very good at cracking passwords.)

 

[Cogman]

What about a quantum computer? Would that be able to brute froce a password much faster?

 

[NikolaeVarius]

Yes. As processing speed goes up, the amount of tries it can do in a second goes up.

 

[f95toli]

Originally posted by: Cogman
What about a quantum computer? Would that be able to brute froce a password much faster?

Probably not. AFAIK there is no know nalgorithm that would allow for a faster solution using a QC.

It is important to remember that a QC is NOT a "generic parallell computer" as such, unless you can find an algorithm which somehow can utilize entanglement (and there are only a few known algorithms) a QC is useless.

 

[her34]

Originally posted by: blackllotus
The supercomputer entry on wikipedia suggests that no supercomputer has achieved computing speeds greater than 1 petaflop (no general purpose one has achieved above 360 teraflops). I think it is to safe to say that testing a password string takes longer than a floating point operation so we can use 2^50 password tests per second as an upper bound.

Now we can calculate how long a password string must be for it to take a year for supercomputer that can test 2^50 passwords per second to crack it. This hypothetical supercomputer can test 2^50 * 60 * 60 * 24 * 365 passwords per year, so with 62 different characters (26 lowercase + 26 uppercase + 10 digits) you should need about log(2^50*60*60*24*365)/log(62) = 12.7 -> 13 characters.

EDIT: Guess you don't need this anymore...

although the current best is 360teraflops, from the wiki page it seems that peak speed is doubling almost every other year.

for logging into somewhere this doesn't really matter because password attempts are limited but encrypted files are susceptible to bruteforce, a problem that can happen when laptops are stolen.




will passwords become pointless protection in 50-100 years, like wep is today?

 

[DrPizza]

I've never figured out why people hate longer, more secure passwords... Pick a pattern on the keyboard that your fingers can type very quickly, practice it a few times to make sure you can hit 8 characters in under a second... sorta like this:
1w7dfikl; 1w7dfikl; 1w7dfikl; 1wd7dfilk; (that's 9 characters, I used a thumb)
See where those characters are on your keyboard... if your hands are the same size as mine, turning your hands at a slight angle makes the keys just sort of fit just right under your fingers.
Then, if you're really paranoid, create a different password for every site you go to by simply appending one more letter to the end, i.e. for anandtech, your password becomes 1w7dfilkl;A For gmail, the A becomes a G, etc.

 

[smack Down]

It is rare that a truely brute force attack is done. An attack requires the password file. Once the hackers have the password file they can run an attack on it. Having the file allows you to test the guessed password against any account of the system which greatly increase the chance of success. The attackers use intelligent guess like trying every word with a number at the end.

Longer password are useless. Any system should lock the user out after a bunch of wrong guess and delay accepting new guess for after a wrong guess to make brute force a completely worthless attack vector unless you have the password file. If an attacker gets the password file that is because the administrators suck and the box has already been rooted so getting user passwords isn't important anyways. Part of fixing a rooted box would be to change all users passwords.

Strong passwords are academic security and not real.

 

[oldman420]

i think that systems should self destruct the data in the file or obliviate the user account after so many tries.

 

[Gannon]

Originally posted by: her34
edit: thanks for responses. i'd generally consider 100 years impractical



i read somewhere how the ever increasing power of computers is causing the length of a "safe" password to increase beyond what typical people would want to deal with (less than 10 characters)

using uppercase, lowercase, and numbers, how long would the password have to be to make bruteforce impractical by anyone (person, organization, etc)?

or does anyone know how many passwords/second the world's fastest supercomputer would be able to calculate?

There is always a solution to brute force attacks for many applications, especially when you are required to login to live networks: Monitoring network/login behavior, and limiting the amount of login attempts before a ban or account lock.

Although I agree that passwords will become increasingly more useless but... you could simply have two or three passwords, instead of requiring one. (think more then one lock on a door). With applications like AI roboform (google roboform), I think it is do-able, since once you save your strange password it's "Fire and forget" and the password is stored in files on your computer. Although if you ever lost them or didn't back them up or keep them somewhere safe, it'd be hell if you ever suffered a crash or lost password.

 

[Skud]

I don't understand why people don't use pass PHRASES more often.. I can type it "Gimme3 s0me Love!" faster than I can type in some computer-generated random sequence and it's easier to remember to boot..

I use a pass phrase whenever I can..

Riley

 

[CycloWizard]

Originally posted by: oldman420
i think that systems should self destruct the data in the file or obliviate the user account after so many tries.

Great, but what happens when you piss off the neighborhood kid and he decides that he'll nuke all your files using the security method you just suggested?

 

[Karot]

Just a reminder that the US Government would allow you to use any sort of encryption that they can't crack themselves.

 

[Matt1970]

Most places will lock you out after 3 failed log-in attempts. You are also limited to the amount of log-in attempts per minute or whatever as the connection to that computer can handle.

 

[TanisHalfElven]

i think the OP is refering to encrypted files not login systems.

now one thing of consider is a lot of encryption mechanisms employ hashing the password and using that as the encryption key. so while you typing a small password it gets hashed into a long number and that is the real password. so to decypt the data you'd need the hash which is insanely long.

 

[RaynorWolfcastle]

Originally posted by: tanishalfelven
i think the OP is refering to encrypted files not login systems.

now one thing of consider is a lot of encryption mechanisms employ hashing the password and using that as the encryption key. so while you typing a small password it gets hashed into a long number and that is the real password. so to decypt the data you'd need the hash which is insanely long.

True enough, but generally the hashing function is known by the attacker (much like the encryption algorithm is known to the attacker); so the hashing just adds a bit of time as the attacker must generate the hash for all passwords shorter than 10 letters.

Since hashes are 1-to-1 functions you don't really increase the search space...

 

[CycloWizard]

Originally posted by: Karot
Just a reminder that the US Government would allow you to use any sort of encryption that they can't crack themselves.

Spoken like someone that has never worked for the government. (I assume that you meant 'would NOT allow')

 

[TanisHalfElven]

Originally posted by: RaynorWolfcastle

Originally posted by: tanishalfelven
i think the OP is refering to encrypted files not login systems.

now one thing of consider is a lot of encryption mechanisms employ hashing the password and using that as the encryption key. so while you typing a small password it gets hashed into a long number and that is the real password. so to decypt the data you'd need the hash which is insanely long.

True enough, but generally the hashing function is known by the attacker (much like the encryption algorithm is known to the attacker); so the hashing just adds a bit of time as the attacker must generate the hash for all passwords shorter than 10 letters.

Since hashes are 1-to-1 functions you don't really increase the search space...

ever heard of salt.

 

[RaynorWolfcastle]

Originally posted by: tanishalfelven

Originally posted by: RaynorWolfcastle

Originally posted by: tanishalfelven
i think the OP is refering to encrypted files not login systems.

now one thing of consider is a lot of encryption mechanisms employ hashing the password and using that as the encryption key. so while you typing a small password it gets hashed into a long number and that is the real password. so to decypt the data you'd need the hash which is insanely long.

True enough, but generally the hashing function is known by the attacker (much like the encryption algorithm is known to the attacker); so the hashing just adds a bit of time as the attacker must generate the hash for all passwords shorter than 10 letters.

Since hashes are 1-to-1 functions you don't really increase the search space...

ever heard of salt.

No actually, but I just looked it up. It makes sense, I guess I should have thought of it before posting (or at least looked it up ).

 

[aurareturn]

I don't get why people don't put spaces in their password. For example, having the password "ap ple" is a lot harder to crack than "apple".

 

[her209]

This is why most systems only allow you to attempt to login after so many times before it blocks any additional attempts.

 

[Skeeedunt]

Originally posted by: tanishalfelven

Originally posted by: RaynorWolfcastle

Originally posted by: tanishalfelven
i think the OP is refering to encrypted files not login systems.

now one thing of consider is a lot of encryption mechanisms employ hashing the password and using that as the encryption key. so while you typing a small password it gets hashed into a long number and that is the real password. so to decypt the data you'd need the hash which is insanely long.

True enough, but generally the hashing function is known by the attacker (much like the encryption algorithm is known to the attacker); so the hashing just adds a bit of time as the attacker must generate the hash for all passwords shorter than 10 letters.

Since hashes are 1-to-1 functions you don't really increase the search space...

ever heard of salt.

I thought salt was just to make lookup tables impractical? My understanding was that the salt was generally known by the attacker, and thus irrelevant in a brute-force attack.