The following sentences can lead to misunderstandings:
"If the salt argument is not provided, one will be randomly generated by PHP each time you call this function."
Comment: false for DES, true for MD5
...
"If you are using the supplied salt, you should be aware that the salt is generated once. If you are calling this function repeatedly, this may impact both appearance and security."
Comment: true for DES, false for MD5
Two sentences nearby, and they are telling incompatible stories. At least I was confused for quit a time.
My analysis after some tests:
If the parameter 'salt' is omitted, there are two possibilities:
* if the default encryption is Standard DES, there is a static salt provided. If you call crypt() with the same password again, the salt and the hash is the same. Effectively, there is no salt functionality. (untested)
* if the default encryption is crypt MD5, always random salt is added, and multiple calls of crypt with the same password provide different salt and hashes. I have tested this to be true.
Experiment (PHP version: 4.4.0 SuSE 10 and PHP version: 5.1.2 SuSE 11):
<?php
echo 'result: ' . crypt('somepassword');
echo 'result: ' . crypt('somepassword');
echo 'result: ' . crypt('somepassword');
?>
result: $1$K2D8DGwq$b05uO37aMwO4rnDlB9Rsi1
result: $1$aPBvu2y.$213YVEs8/5m.jMCXSScly/
result: $1$dW3Xu2p6$nuCtJe2zzlgBMLxN2oZCx/
I hope, i can prevent some headache.
crypt
説明
string crypt ( string str, string [salt] )crypt()は、Unix 標準の DES 暗号を使って文字列を暗号化します。引数には暗号化したい文字列を指 定し、オプションで暗号化のベースとなる 2 文字の'salt' 文字列を指 定します。暗号化に関する関数の詳細は Unix man ページを参照してく ださい。
salt引数が指定されない場合、この関数をコールするたびに PHPによりランダムに生成されます。
いくつかのオペレーティングシステムは、複数の暗号化方式をサポート しています。実際、標準のDES暗号の替わりにMD5に基づく暗号化アルゴ リズムが使用されることが時々あります。暗号化手法は、salt引数によ り選定されます。インストール時に、PHPはcrypt関数の機能を定義し、 他の暗号化方式用のsaltを受付ます。saltが指定されない場合、システ ムのデフォルトの暗号化手法がMD5でない限りPHPはデフォルトで標準の2 文字DES saltを自動的に生成します。暗号化手法がMD5の場合、MD5互換 のsaltが生成されます。PHPは、定数CRYPT_SALT_LENGTH を設定します。 この定数により、標準の2文字のsaltまたはより長い12文字MD5 saltのど ちらが適用可能であるかを知ることが可能です。
生成されたsaltを使用する場合、saltが生成されるのは一回限りである ことに気付く必要があります。この関数を繰り返しコールする場合、こ のことは、出力だけでなく、セキュリティにも影響を与える可能性があ ります。
標準DES暗号crypt()は、出力の最初の2文字をsalt として使用します。 strの最初の8文字を使用するだけです。 このため、(同じsaltを使用した場合、)同じ8文字から始まるより長い文 字列について同じ結果が生成されます。
crypt() 関数が複数の暗号化手法をサポートするシステムにおいては、 指定した型を利用可能であるかにより、次の定数を0又は1に設定します。
CRYPT_STD_DES - 2文字 SALT の標準 DES 暗号
CRYPT_EXT_DES - 9文字の拡張DES暗号
CRYPT_MD5 - $1$ で始まる12文字のSALTを有するMD5暗号
CRYPT_BLOWFISH - $2$ もしくは $2a$ で始まる16文字の拡張DES暗号
注意: crypt() は単方向アルゴリズムを使用している ため、復号化するための関数 (decrypt) はありません。
例 3. 異なる暗号化手法を用いた crypt() の例
上の例の出力は、たとえば 以下のようになります。
|
md5()およびmcrypt拡 張モジュールも参照ください。
crypt
Marcus dot Kabele at Kabele dot at
20-Jan-2007 04:19
20-Jan-2007 04:19
tserong at sgi dot com
14-Dec-2006 05:18
14-Dec-2006 05:18
Blowfish doesn't use a sixteen character salt, it uses sixteen *bytes* of salt. So (courtesy of the docs for the Crypt::Eksblowfish::Bcrypt Perl module), it's:
"$2", optional "a", "$", two digits, "$", and 22 base 64 digits
If the salt is not long enough, crypt will return "*0" and you will have no idea what is wrong. Interestingly, the example in the documentation with a trailing '$' in the salt does not work. Replace the '$' with a '.', and the output appears as advertised.
picolobo at pilab dot dyndns dot org
22-Sep-2006 05:49
22-Sep-2006 05:49
I had problems with ENCRYPT MySQL function when i tried to compare with the encrypted password (with ENCRYPT).
Another solution i read from "UNIX Advanced programming" where i found about the UNIX system call "crypt()":
Password="tB" //The two first letters of encrypted password
SELECT password from users where Password=ENCRYPT('".$_POST['password']."',Password)
mysql> select password from users where password=encrypt('pasword','tB');
+---------------+
| password |
+---------------+
| tBY8OVuabSiTU |
+---------------+
1 row in set (0.01 sec)
Bye.
> topace at lightbox dot org
> 22-Sep-2005 06:34
>
> To authenticate against a stored crypt in MySQL, simply use:
>
> SELECT ................
> AND Password=ENCRYPT('".$_POST['password']."',Password)
solar at openwall dot com
24-Dec-2005 04:20
24-Dec-2005 04:20
With different password hashing methods supported on different systems and with the need to generate salts with your own PHP code in order to use the more advanced / more secure methods, it takes special knowledge to use crypt() optimally, producing strong password hashes. Other message digest / hashing functions supported by PHP, such as md5() and sha1(), are really no good for password hashing if used naively, resulting in hashes which may be brute-forced at rates much higher than those possible for hashes produced by crypt().
I have implemented a PHP password hashing framework (in PHP, tested with all of PHP 3, 4, and 5) which hides the complexity from your PHP applications (no need for you to worry about salts, etc.), yet does things in almost the best way possible given the constraints of the available functions. The homepage for the framework is:
http://www.openwall.com/phpass/
I have placed this code in the public domain, so there are no copyrights or licensing restrictions to worry about.
P.S. I have 10 years of experience in password (in)security and I've developed several other password security tools and libraries. So most people can feel confident they're getting this done better by using my framework than they could have done it on their own.
hotdog (at) gmx (dot) net
17-Nov-2005 12:34
17-Nov-2005 12:34
WRONG:
$mypassword = "toto";
$smd5_pass = "{SMD5}......." // in openldap
if (preg_match ("/{SMD5}/i", $smd5_pass))
{
$encrypted = substr($md5_pass, 6);
$hash = base64_decode($encrypted);
$salt = substr($hash,16);
$mhashed = mhash(MHASH_MD5, $mypassword . $salt) ;
$without_salt = explode($salt,$hash_hex);
if ($without_salt[0] == $mhashed) {
echo "Password verified <br>";
} else {
echo "Password Not verified<br>";
}
}
$without_salt = explode($salt,$hash_hex); should be $without_salt = explode($salt,$hash);
RIGHT:
$mypassword = "toto";
$smd5_pass = "{SMD5}......." // in openldap
if (preg_match ("/{SMD5}/i", $smd5_pass))
{
$encrypted = substr($md5_pass, 6);
$hash = base64_decode($encrypted);
$salt = substr($hash,16);
$mhashed = mhash(MHASH_MD5, $mypassword . $salt) ;
$without_salt = explode($salt,$hash);
if ($without_salt[0] == $mhashed) {
echo "Password verified <br>";
} else {
echo "Password Not verified<br>";
}
}
bjorninges dot spam at gmail dot com
07-Nov-2005 07:16
07-Nov-2005 07:16
Note to topace's code: you should not use $_POST['password'] directly in your query as you are open to sql-injects.
Use the quote_smart() function from http://no.php.net/mysql_real_escape_string before adding user-submitted data to the query
topace at lightbox dot org
23-Sep-2005 01:34
23-Sep-2005 01:34
To authenticate against a stored crypt in MySQL, simply use:
SELECT ................
AND Password=ENCRYPT('".$_POST['password']."',Password)
MagicalTux at FF dot st
15-Jun-2005 04:53
15-Jun-2005 04:53
Since many of you are wondering why when providing salt the characters over the 8th are ignored in the password, I'll clarify it a bit.
By default, PHP will try to use the best encryption method available on your system : MD5 or Simple DES.
Usually this is the MD5 method ($1$).
In this case, the salt must look like : "$1$xxxxxxxx$" where x are random ASCII characters. When you use MD5 passwords, all characters of the password are encrypted in this 34 characters hash.
However if your salt starts with an ASCII character, the system will assume it's a standard DES encrypted password. The main weakness of this system : only the 8 first characters of the password are used.
A correct version of the code :
<?php
function makesalt($type=CRYPT_SALT_LENGTH) {
switch($type) {
case 8:
$saltlen=9; $saltprefix='$1$'; $saltsuffix='$'; break;
case 2:
default: // by default, fall back on Standard DES (should work everywhere)
$saltlen=2; $saltprefix=''; $saltsuffix=''; break;
#
}
$salt='';
while(strlen($salt)<$saltlen) $salt.=chr(rand(64,126));
return $saltprefix.$salt.$saltsuffix;
}
$salt=makesalt();
$longpassword='fez1c89ez1c98ez4c89z4eqf98ez';
$encrypted = crypt($longpassword, $salt);
$encrypted2 = crypt(substr($longpassword, 0, 8), $encrypted);
if ($encrypted == $encrypted2) {
echo 'Match: Weak encryption method (Standard DES)';
} else {
echo 'NoMatch: Strong encryption method (MD5)';
}
?>
If you use makesalt(2) you will force usage of Standard DES method, and the passwords will match. If you just use makesalt() there's great chances you'll have a MD5 password (don't know any system used nowadays which does not support MD5 passwords).
Finally, do not look at md5() PHP function if you want a md5 password, that's not related. UNIX MD5 passwords uses a salt, are 34 character long and start with $1$. The reply from md5() is 32 characters long, and is more adapted for file integrity check (call that a checksum).
Some people use sha1() passwords but without salt. Here's my simple sha1crypt function which will work with standard crypt passwords (by calling crypt()) AND a home-made "sha1" encryption method.
<?php
// NOTE: This function requires PHP 5.0.0 as we use "raw output" option of sha1()
function sha1crypt($password, $salt=null) {
if ( (is_null($salt)) || (strlen($salt)<1) ) {
$salt='';
while(strlen($salt)<10) $salt.=chr(rand(64,126));
$salt='$sha1$'.$salt.'$';
}
if ($salt{0}!='$') return crypt($password, $salt);
$tmp=explode('$',$salt);
if ($tmp[1]!='sha1') return crypt($password, $salt);
$saltstr=$tmp[2];
if (strlen($saltstr) != 10) return crypt($password, $salt);
$encrypt=base64_encode(sha1($saltstr.$password,true));
return '$sha1$'.$saltstr.'$'.$encrypt;
}
// without salt, sha1crypt() will generate on
$pass=sha1crypt('foobar');
echo $pass."\n";
// pass directly password as salt - different output as password is not the same
echo sha1crypt('foobarbaz',$pass)."\n";
// same password - same output
echo sha1crypt('foobar',$pass)."\n";
// Encrypt using MD5 passwords
echo sha1crypt('foobar','$1$blahblahg$')."\n";
?>
will output:
$sha1$oFkYeI|vuu$d3n7D30OnecZSbS6KIbxCch608A=
$sha1$oFkYeI|vuu$iA8KmbCZun1G1gEw2qVr42ELVH4=
$sha1$oFkYeI|vuu$d3n7D30OnecZSbS6KIbxCch608A=
$1$blahblah$/8Hme91aEkHzLaVk0g9EQ0
My sha1-encrypted passwords are 45 characters long.
Remember to read that too before using SHA1 passwords too :
http://www.schneier.com/blog/archives/2005/02/sha1_broken.html
ceo at l-i-e dot com
07-Jun-2005 10:03
07-Jun-2005 10:03
The people confused about the first 2 characters of the plain-text password being used as the salt are... confused.
AFTER you crypt() your plain-text password, with 2 RANDOM characters, the RESULTING scrambled output with have the 2 randomly-selected characters as its first 2 characters.
Later, to check a password presented by the user, you use the first 2 characters of the SCRAMBLED password as the salt.
So you do not need to remember the salt elsewhere -- it is buried in the encrypted output.
This is by design by very very very clever cryptologists, and in no way, shape, or form decreases the "security" of the algorithm.
The same is true of all the encryption algorithms here -- The salt may be 27 characters long, and embedded in the middle of the crypted string, but it's still there, and that's Good.
I hope this note decreases the confusion caused by the 2 early notes discussing salt.
Vlad Alexa Mancini mancin at nextcode dot org
16-May-2005 05:57
16-May-2005 05:57
cleaner version of shadow() and with more ascii chars
<?php
function shadow ($input){
for ($n = 0; $n < 9; $n++){
$s .= chr(rand(64,126));
}
$seed = "$1$".$s."$";
$return = crypt($input,$seed);
return $return;
}
>
kevin at sylandroXgetridofthisbitX dot com
20-Apr-2005 04:55
20-Apr-2005 04:55
Even worse, using the first two letters of the password as the salt removes the whole point of having salt in the first place, which is that without it, it's trivial to create a dictionary of passwords-to-encrypted-passwords. Without salt (or with using a salt directly derived from the password) there's only one possible encrypted password per password. With the two character salt, there's 4096, which makes the idea a lot less feasible.
19-Apr-2005 05:57
>icecube at fr dot fm 07-Mar-2002 09:53
>To generate a .htaccess-based authentication with DES,
>you have to use the first two characters of your password
>as salt.
This is a _really_ bad idea, seeing how the two first letters of your password will be reveiled, effectively reducing the quality of your password.
Your example assigns "sonH/h2hpGtHk" to $ht_pass, where "so" of course are the first to letters in something. Try it out for yourself. Switching the password to "icecube" gives "ickJ.ZxzjXpNE" as the output. Get my point?
The salt is always included in the encrypted string. Use a random salt instead.
postal2600 at yahoo dot com
16-Mar-2005 07:07
16-Mar-2005 07:07
I'll take the example from above and take a situation that I've ecountered and show you something that puts this function in a dark shadow of insecurity:
<?php
$password = crypt('postal2600','CyberBoard');
if (crypt('postal26??', $password) == $password) {
echo "Password verified!";
}
?>
Instad of ?? you can put anything an see that allway the password will be verified. So i strongly recomand that the md5 function should be used instead.
kb at -NOSPAM-diggersdive dot info
31-Dec-2004 01:18
31-Dec-2004 01:18
It is likely that the php version of crypt() interfaces relatively directly with the C-library crypt() function, usually contained in libcrypt.
Depending on your implementation and the default salt, you may or may not get the same crypt'd password out of eight characters as you would a longer string starting with those same eight characters.
See crypt(3) on your system for more documentation on how it works based on different encryption types.
genius at clan-aftershock dot com
29-Nov-2004 10:10
29-Nov-2004 10:10
I discovered in the script stated before, when using a salt string the passwords length cannot be longer than 8 characters. However, when not using a salt string, it can exceded that limit.
Modified Example from athony, not using the salt.
<?php
$password = "qwertyuiopasdfghjkl";
// Not using any salts
$encrypted = crypt($password);
// Now do the comparison
$shortPass = substr($password, 0, 8);
if (crypt( $shortPass, $encrypted ) == $encrypted )
echo "The passwords match";
else
echo "The passwords do not match";
?>
This will print: "The passwords do not match"
The check for password would still work if it was correct. Just by shortening the password string to eight, or extending the substr, or decalring the shortpass the same as the pass, it would be: "The passwords do match"
So the eight letter limit only works when using a salt.
antony at anonymous dot anon dot com
27-Nov-2004 01:21
27-Nov-2004 01:21
There appears to be a limitation with this function, where it only validates upto a characters, therefore the 9th character onwards can be ommitted, which limits useful passwords to 8 characters only. Example
<?php
$password = "qwertyuiopasdfghjkl";
// Encrypt the password, but let's use a known MD5 hash as the salt
$salt = ""0f2d92cee71e5f93f3abecdc666a6b7d";
$salt = substr($salt, 0, CRYPT_SALT_LENGTH );
$encrypted = crypt($password, $salt);
// Now do the comparison
$shortPass = substr($password, 0, 8);
if (crypt( $shortPass, $encrypted ) == $encrypted )
echo "The passwords match";
else
echo "The passwords do not match";
?>
This will print: "The passwords match" even though $shortPass is "qwertyui" and $password is "qwertyuiopasdfghjkl"
thorhajo at gmail dot com
03-Sep-2004 07:34
03-Sep-2004 07:34
Here's a little function I wrote to generate MD5 password hashes in the format they're found in /etc/shadow:
function shadow($password)
{
$hash = '';
for($i=0;$i<8;$i++)
{
$j = mt_rand(0,53);
if($j<26)$hash .= chr(rand(65,90));
else if($j<52)$hash .= chr(rand(97,122));
else if($j<53)$hash .= '.';
else $hash .= '/';
}
return crypt($password,'$1$'.$hash.'$');
}
I've written this so that each character in the a-zA-Z./ set has a 1/54 of a chance of being selected (26 + 26 + 2 = 54), thus being statistically even.
aidan at php dot net
05-Jul-2004 11:52
05-Jul-2004 11:52
Text_Password allows one to create pronounceable and unpronounceable passwords.
http://pear.php.net/package/text_password
php at SPAM_tlarson dot com
26-Jun-2002 02:17
26-Jun-2002 02:17
There's always been a bit of confusion as to what makes a good salt and what doesn't. Remember that it doesn't matter at all how easy a salt is to guess. No one ever HAS to guess the salt: it's already given.
The only only important consideration when generating a salt is to make sure that all salts are unique--that way the same password will be encrypted differently (i.e. the encrypted passwords will look different) for different users.
One of the simplest ways to generate a unique salt is to use some string that will be different every time the procedure is called. Here's a simple example:
<?php
$jumble = md5(time() . getmypid());
$salt = substr($jumble,0,$salt_length);
?>
Given a string consisting of the current time (in seconds) concatinated with the current process id, the string will never be the same twice, assuming that the function is never called more than once per second. Calculating the md5 sum over that string creates another string from which you can extract any substring and still end up with a unique sequence.
If you're going to be generating more than one password per second, just throw a rand($x,$y) in there to add a little more entropy.