From: unruh on
On 2010-03-02, Phoenix <ribeiroalvo(a)gmail.com> wrote:
> On 1 Mar, 14:35, Ertugrul S?ylemez <e...(a)ertes.de> wrote:
>
>> Yes, and also note that increasing the length of a password has much
>> greater effect than increasing the size of the alphabet.
>
> No.
>
> In exhaustive search, increasing the length of a password has the same
> effect than increasing the size of the alphabet.

ln(N)=L*ln(A)
where A is the alphabet, and L is the length of the password. Ie, L has
a far bigger effect on the number of passwords than does a similar
increase in password.
>
> See:
> For example:
>
> 64^8 is the sane as 256^6

So increasing the password length by 30% is the same as increasing the
alphabet by a factor of 4 (400%)


>
> A 64 letter alphabet and pw_length=8, is the same as an 256 letter
> alphabet and pw_length=6
>
> 64^8=2^(6*8)
> 256^6=2^(8*6)
>
> Paulo Marques, thanks for your great posts.
From: Phoenix on
On 2 Mar, 01:07, unruh <un...(a)wormhole.physics.ubc.ca> wrote:

> ln(N)=L*ln(A)
> where A is the alphabet, and L is the length of the password. Ie, L has
> a far bigger effect on the number of passwords than does a similar
> increase in password.
> So increasing the password length by 30% is the same as increasing the
> alphabet by a factor of 4 (400%)
>
>

unruh, imagine that we have two alphabets:

A1 with n characters
A2 with 2n characters

and

Length_pw1=Length_pw2

It's easy to see, than costs more to find pw2 then pw1.
At the end, what matters is the result of 2^(A*L).

Thats why, if we talk in bits, we need more then if we talk about
Bytes or Hexa.
Binary have A=2 and Bytes have A=8 and Hexa have A=16.

The importance of a bigger alphabet is the need of smaller length_pw,
for the same safety and, consequently, easy to humans memorize it.
I think you agree with me, then is more easy to memorize i.e 2 letters
in a 256 alphabet, then i.e. 16 letters in an 32 alphabet.
From: J.D. on
> It's easy to see, than costs more to find pw2 then pw1.
> At the end, what matters is the result of 2^(A*L).

Uh, don't you mean A^L? Or perhaps 2^(log2(A)*L)?

From: unruh on
On 2010-03-02, Phoenix <ribeiroalvo(a)gmail.com> wrote:
> On 2 Mar, 01:07, unruh <un...(a)wormhole.physics.ubc.ca> wrote:
>
>> ln(N)=L*ln(A)
>> where A is the alphabet, and L is the length of the password. Ie, L has
>> a far bigger effect on the number of passwords than does a similar
>> increase in password.
>> So increasing the password length by 30% is the same as increasing the
>> alphabet by a factor of 4 (400%)
>>
>>
>
> unruh, imagine that we have two alphabets:
>
> A1 with n characters
> A2 with 2n characters
>
> and
>
> Length_pw1=Length_pw2
>
> It's easy to see, than costs more to find pw2 then pw1.
> At the end, what matters is the result of 2^(A*L).

??? What is A. The answer is 2^(ln_2(n)*L) or n^L

>
> Thats why, if we talk in bits, we need more then if we talk about
> Bytes or Hexa.
> Binary have A=2 and Bytes have A=8 and Hexa have A=16.
>
> The importance of a bigger alphabet is the need of smaller length_pw,
> for the same safety and, consequently, easy to humans memorize it.
> I think you agree with me, then is more easy to memorize i.e 2 letters
> in a 256 alphabet, then i.e. 16 letters in an 32 alphabet.

Is it easier to memorize a hex number than a decimal? Not if you are
used to decimal numbers. The "most efficient" alphabet has about 3
characters (least stuff to memorize), if I remember correctly.
A bigger alphabet costs far more to memorize the alphabet than you gain
from fewer characters in the word.
That is because of the ln(n) in the above.



From: Phoenix on
On 2 Mar, 12:38, Ertugrul Söylemez <e...(a)ertes.de> wrote:

> Note that you don't memorize alphabets, but passwords.  Shorter
> passwords in exchange for larger alphabets are generally easier to
> memorize, that's why we use dictionary-based passwords like Diceware.
>
> The tradeoff is:  If we want to decrease the length needed, we must make
> our alphabets rather huge.  Just adding more ASCII characters won't be
> enough.  Instead our alphabet must become a dictionary.


Completly agree.