RE: minimum password length check
On 18 Dec 2009 at 1:13, shyam_i...@dell.com wrote:
-Original Message-
From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com]
On Behalf Of Ulrich Windl
Sent: Wednesday, December 16, 2009 1:08 PM
To: open-iscsi@googlegroups.com
Subject: Re: minimum password length check
On 15 Dec 2009 at 22:47, shyam_i...@dell.com wrote:
From the spec:
CHAP secrets MUST be an integral number of bytes (octets). A
compliant implementation SHOULD NOT continue with the login step
in
which it should send a CHAP response (CHAP_R, Section 11.1.4
Challenge Handshake Authentication Protocol (CHAP)) unless it can
verify that the CHAP secret is at least 96 bits, or that IPsec
encryption is being used to protect the connection.
You picked up an interesting issue: The Microsoft Initiator limits
the length of
the secret to 16 characters (AFAIR). I wrote a lottle program that
generates
random secrets and estimated the entropy (i.e. number of bits):
With 16 random letters, you are at about 92 bits (e.g.
mMPuhxfKAYuIFTjZ)
With 16 random letters with digits you are at about 95 bits (e.g.
b3v4B8mRoiFWjpF9)
What algorithm are you using to arrive at this ...
Googling(and some of my information theory lit..) almost always hints me
to shannon's theorem to find the randomness of a character string ...
As pointed out before, this is the randomness of a string the program creates
itself. So if you unly use the set {A,B,C,D} that two bits per randomly chosen
element. Then a ten-character string will have 10*2=20 bits of entropy. More
complex scenarios are similar.
Now if you have the String ABCD its randomness will actually vary, depending
on
the range of characters chosen. For the full alphabet an 'A' will have more
significant bits than in the example above.
IMHO that's the problem: If you don't know the range being used, you cannot
guess
the entropy. Of course you can try to derive the range from the string being
seen,
but that's just a guess.
Likewise, when considering December as a password, it's true entropy is much
less once you know that the range of passwords are only month names. (And so on)
Check this
http://www.redkestrel.co.uk/Articles/RandomPasswordStrength.html
Yes that's the basics, but you never know how randomly the user picked it's
characters. Obviously this wasn't very random, even though a program might
think
so:
qwertzuiop
Regards,
Ulrich
P.S: Off-topic, isn't it?
--
You received this message because you are subscribed to the Google Groups
open-iscsi group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
-Original Message-
From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com]
On Behalf Of Ulrich Windl
Sent: Friday, December 18, 2009 3:24 PM
To: open-iscsi@googlegroups.com
Subject: RE: minimum password length check
On 18 Dec 2009 at 1:13, shyam_i...@dell.com wrote:
-Original Message-
From: open-iscsi@googlegroups.com [mailto:open-
is...@googlegroups.com]
On Behalf Of Ulrich Windl
Sent: Wednesday, December 16, 2009 1:08 PM
To: open-iscsi@googlegroups.com
Subject: Re: minimum password length check
On 15 Dec 2009 at 22:47, shyam_i...@dell.com wrote:
From the spec:
CHAP secrets MUST be an integral number of bytes (octets). A
compliant implementation SHOULD NOT continue with the login
step
in
which it should send a CHAP response (CHAP_R, Section 11.1.4
Challenge Handshake Authentication Protocol (CHAP)) unless it
can
verify that the CHAP secret is at least 96 bits, or that
IPsec
encryption is being used to protect the connection.
You picked up an interesting issue: The Microsoft Initiator
limits
the length of
the secret to 16 characters (AFAIR). I wrote a lottle program that
generates
random secrets and estimated the entropy (i.e. number of bits):
With 16 random letters, you are at about 92 bits (e.g.
mMPuhxfKAYuIFTjZ)
With 16 random letters with digits you are at about 95 bits (e.g.
b3v4B8mRoiFWjpF9)
What algorithm are you using to arrive at this ...
Googling(and some of my information theory lit..) almost always
hints
me
to shannon's theorem to find the randomness of a character string
...
As pointed out before, this is the randomness of a string the program
creates
itself. So if you unly use the set {A,B,C,D} that two bits per
randomly
chosen
element. Then a ten-character string will have 10*2=20 bits of
entropy.
More
complex scenarios are similar.
Now if you have the String ABCD its randomness will actually vary,
depending on
the range of characters chosen. For the full alphabet an 'A' will have
more
significant bits than in the example above.
IMHO that's the problem: If you don't know the range being used, you
cannot guess
the entropy. Of course you can try to derive the range from the string
being seen,
but that's just a guess.
Likewise, when considering December as a password, it's true entropy
is much
less once you know that the range of passwords are only month names.
(And so on)
Check this
http://www.redkestrel.co.uk/Articles/RandomPasswordStrength.html
Yes that's the basics, but you never know how randomly the user picked
it's
characters. Obviously this wasn't very random, even though a program
might think
so:
qwertzuiop
Regards,
Ulrich
P.S: Off-topic, isn't it?
I am increasing also beginning to infer both through discussions here
and(internally) that fixing a minimum randomness for a password is not
right in the open-iscsi
I guess this discussion might just offer some insight into the
randomness of a password for a user to set.
This can't just be imposed on an implementation verbatim.
So, if there were an implementation Must and a deployment Must, this
falls under the latter.
I guess the 16 character guideline in the Microsoft initiator is more of
a guideline so that users can remember them and not write them in
post-its to compromise the secret further ... :)
-Shyam
--
You received this message because you are subscribed to the Google Groups
open-iscsi group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
-Original Message- From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com] On Behalf Of Ulrich Windl Sent: Thursday, December 17, 2009 1:27 PM To: open-iscsi@googlegroups.com Subject: RE: minimum password length check On 17 Dec 2009 at 0:55, shyam_i...@dell.com wrote: Essentially what you are saying is that we haven't implemented the secret's bit randomness calculation to check if has atleast 96bits of entropy. No, I just wanted to point out that the quality of a secret key cannot simply be measured with strlen(password), and that 96 bits of randomness may require a longer string as one might initially have guessed. Right I get you right then.. Don't want to rework on getting the entropy of the secret. Do you mind open-sourcing it so we could do the checks as I detailed in the previous mail? Thanks, Shyam -- You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-is...@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
On 17 Dec 2009 at 16:08, shyam_i...@dell.com wrote: -Original Message- From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com] On Behalf Of Ulrich Windl Sent: Thursday, December 17, 2009 1:27 PM To: open-iscsi@googlegroups.com Subject: RE: minimum password length check On 17 Dec 2009 at 0:55, shyam_i...@dell.com wrote: Essentially what you are saying is that we haven't implemented the secret's bit randomness calculation to check if has atleast 96bits of entropy. No, I just wanted to point out that the quality of a secret key cannot simply be measured with strlen(password), and that 96 bits of randomness may require a longer string as one might initially have guessed. Right I get you right then.. Don't want to rework on getting the entropy of the secret. Do you mind open-sourcing it so we could do the checks as I detailed in the previous mail? I don't have the code you are looking for, because what I have is Perl, and it does _create_ random secrets following a pattern, optionally outputting the estimated bits of randomness. Originally written to create similar, but different, not very obvious root passwords for a set of similar machines. Something completely different... Regards, Ulrich -- You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-is...@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
-Original Message- From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com] On Behalf Of Ulrich Windl Sent: Wednesday, December 16, 2009 1:08 PM To: open-iscsi@googlegroups.com Subject: Re: minimum password length check On 15 Dec 2009 at 22:47, shyam_i...@dell.com wrote: From the spec: CHAP secrets MUST be an integral number of bytes (octets). A compliant implementation SHOULD NOT continue with the login step in which it should send a CHAP response (CHAP_R, Section 11.1.4 Challenge Handshake Authentication Protocol (CHAP)) unless it can verify that the CHAP secret is at least 96 bits, or that IPsec encryption is being used to protect the connection. You picked up an interesting issue: The Microsoft Initiator limits the length of the secret to 16 characters (AFAIR). I wrote a lottle program that generates random secrets and estimated the entropy (i.e. number of bits): With 16 random letters, you are at about 92 bits (e.g. mMPuhxfKAYuIFTjZ) With 16 random letters with digits you are at about 95 bits (e.g. b3v4B8mRoiFWjpF9) What algorithm are you using to arrive at this ... Googling(and some of my information theory lit..) almost always hints me to shannon's theorem to find the randomness of a character string ... Check this http://www.redkestrel.co.uk/Articles/RandomPasswordStrength.html -- You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-is...@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
-Original Message-
From: open-iscsi@googlegroups.com [mailto:open-is...@googlegroups.com]
On Behalf Of Ulrich Windl
Sent: Wednesday, December 16, 2009 1:08 PM
To: open-iscsi@googlegroups.com
Subject: Re: minimum password length check
On 15 Dec 2009 at 22:47, shyam_i...@dell.com wrote:
From the spec:
CHAP secrets MUST be an integral number of bytes (octets). A
compliant implementation SHOULD NOT continue with the login step
in
which it should send a CHAP response (CHAP_R, Section 11.1.4
Challenge Handshake Authentication Protocol (CHAP)) unless it can
verify that the CHAP secret is at least 96 bits, or that IPsec
encryption is being used to protect the connection.
You picked up an interesting issue: The Microsoft Initiator limits
the length of
the secret to 16 characters (AFAIR). I wrote a lottle program that
generates
random secrets and estimated the entropy (i.e. number of bits):
With 16 random letters, you are at about 92 bits (e.g.
mMPuhxfKAYuIFTjZ)
With 16 random letters with digits you are at about 95 bits (e.g.
b3v4B8mRoiFWjpF9)
The bad thing is that some characters look quite similar so users,
like
'0'
and'O', or '1' and 'l'. When trying to omit those potentially
confusing
characters
(plus adding other punctuation characters, leaving out space for
obvious reasons),
I'm at about 83 bits (e.g. u\FphNwuuWCT74+h).
As a side note: Passwords with only six letters in one case only make
about 28
bits. Now if you think that most users will use words, you can guess
how poor
those passwords actually are.
Using the fully printable ASCII characterset without those characters
that are
considered unsafe in UNIX, 16 characters would have about 102 bits
of
entropy
(e.g. !)Zbl(p7%Hd88LT)
The spec suggests that a chap secret be at least 96bits or (12
characters) but I see that only the AUTH_STR_MAX_LEN of 256
characters
is used for error checking.
Even when just using digits, that would be 850 bits of entropy,
probably enough ;-
)
Regards,
Ulrich
Am I reading this correctly ?
-Shyam Iyer
--
You received this message because you are subscribed to the Google
Groups open-iscsi group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to open-
iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/open-iscsi?hl=en.
Essentially what you are saying is that we haven't implemented the
secret's bit randomness calculation to check if has atleast 96bits of
entropy.
So I guess we should do some thing like this
If (check_96bit_entropy(secret) secret AUTH_MAX_STR_LEN) {
Use_secret
}
else {
Secret not strong enough ..throw error...
}
--
You received this message because you are subscribed to the Google Groups
open-iscsi group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/open-iscsi?hl=en.
Re: minimum password length check
shyam_i...@dell.com wrote:
So I guess we should do some thing like this
If (check_96bit_entropy(secret) secret AUTH_MAX_STR_LEN) {
Use_secret
}
else {
Secret not strong enough ..throw error...
}
We do not check. The only problem would be if we added one now lots of
people are going to get errors in existing set ups. Some might not boot.
Maybe add a error message for a while, then make it mandatory in a later
release.
--
You received this message because you are subscribed to the Google Groups
open-iscsi group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/open-iscsi?hl=en.
RE: minimum password length check
On 17 Dec 2009 at 0:55, shyam_i...@dell.com wrote: Essentially what you are saying is that we haven't implemented the secret's bit randomness calculation to check if has atleast 96bits of entropy. No, I just wanted to point out that the quality of a secret key cannot simply be measured with strlen(password), and that 96 bits of randomness may require a longer string as one might initially have guessed. Regards, Ulrich -- You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-is...@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
