加密存储模型

SSL/SSH 能保护客户端和服务器端交换的数据,但 SSL/SSH 并不能保护数据库中已有的数据。SSL 只是一个加密网络数据流的协议。

如果攻击者取得了直接访问数据库的许可(绕过 web 服务器),敏感数据就可能暴露或者被滥用,除非数据库自己保护了这些信息。对数据库内的数据加密是减少这类风险的有效途径,但是只有很少的数据库提供这些加密功能。

对于这个问题,有一个简单的解决办法,就是创建自己的加密机制,然后把它用在 PHP 程序内。PHP 有几个扩展库可以完成这个工作,比如说 McryptMhash 等,它们包含多种加密运算法则。脚本在插入数据库之前先把数据加密,以后提取出来时再解密。有关加密如何工作的例子请参考相关手册。

对某些真正隐蔽的数据,如果不需要以明文的形式存在(即不用显示),可以考虑用散列算法。使用散列算法最常见的例子就是把密码经过 MD5 加密后的散列存进数据库来代替原来的明文密码。参见 crypt()md5()

例子 27-1. 对密码字段进行散列加密

<?php

// 存储密码散列
$query  = sprintf("INSERT INTO users(name,pwd) VALUES('%s','%s');",
            
addslashes($username), md5($password));
$result = pg_query($connection, $query);

// 发送请求来验证用户密码
$query = sprintf("SELECT 1 FROM users WHERE name='%s' AND pwd='%s';",
            
addslashes($username), md5($password));
$result = pg_query($connection, $query);

if (
pg_num_rows($result) > 0) {
    echo
'Welcome, $username!';
} else {
    echo
'Authentication failed for $username.';
}

?>

add a note add a note User Contributed Notes
Ketos
03-Nov-2006 08:05
You should always hash the password with the username, so store md5($password . $username) instead of md5($password) This way the cracker cannot use a precomputed set of hashes of common keys, but has to recompute the hashes for each user.

Double hashing also can help in some situations as it solves length extension problems with all hash functions. Also you should use SHA-256, as it has a longer hash length and thus it takes 2^128 steps to find a collision
roysimkes at hotmail dot com
17-Aug-2006 08:10
But the thing in double hashing is to increase the time that they can crack your password! A passwod that takes a year to break is quite secure(!)
And even if they learn somehow that you are double hashing your passwords, how will they learn which one you are using first! Instead of using md5() twice try using sha1(md5()) or vice versa. An the breaker should check more things. For only one password he has to check at least 4 different types (md5 only, sha1 only, md5(sha1) and sha1(md5) ) it will take quite a time to break it! (Yeah if he can see your source files it doesn't mean anything. But if they can see that files, you are already dead in the water)
And also put a salt variable in it! If that's random salt, it will take a very long time to crack the password. The best security is to use a hybrid solution. If you add salt to your passwords they have to use your app to crack the code. And if you add some more features like after 3 times password failure add a secret question and a secret answer, a visual confirmation code! Things are getting complicated for the automated script!
Well yes if you want to protect something, do not publish them in the internet. With enough dedication, time and resources every password can be cracked. But if the time is more then years, who will waste time on it? Or the password to be cracked worth that time?
dan[ddot) crowley [att]gmail {dott]com
14-Jul-2006 04:31
A note to the people who think that hashing a password multiple times makes it uncrackable:

It doesn't.

If you're using a standard hash cracker or rainbow tables, sure. But if you've got the smarts to design your own hash cracker, you can just double-hash every string where it would normally be hashed once, and then at best your hashes take twice as long to crack.
Fairydave at the location of dodo.com.au
12-Feb-2006 10:58
I think the best way to have a salt is not to randomly generate one or store a fixed one. Often more than just a password is saved, so use the extra data. Use things like the username, signup date, user ID, anything which is saved in the same table. That way you save on space used by not storing the salt for each user.

Although your method can always be broken if the hacker gets access to your database AND your file, you can make it more difficult. Use different user data depending on random things, the code doesn't need to make sense, just produce the same result each time. For example:

if ((asc(username character 5) > asc(username character 2))
{
   if (month the account created > 6)
     salt = ddmmyyyy of account created date
   else
     salt = yyyyddmm of account created date
}
else
{
   if (day of account created > 15)
     salt = user id * asc(username character 3)
   else
     salt = user id + asc(username character 1) + asc(username character 4)
}

This wont prevent them from reading passwords when they have both database and file access, but it will confuse them and slow them up without much more processing power required to create a random salt
3M
06-Oct-2005 11:46
Be smart and simple!

1. Hash a password once on twice! It's sufficient to use SHA-1 twice! (read no.3)
2. Limit the number of log-in trys per ip! This will be a hard hit for online brute force crackers!
3. The comments below that suggest using a SALT whit MD5 ar good as long as the length of the resulting string (password + SALT) is bigger then the output of the hashing algorithm. For example: SHA-1 has an output of 160 bits or 20 ASCII chars. If the password set has 8 chars then the salt would have to be of 12 chars or longer... the longer the better!
The idea here is that by trying to brute force a hash (second) that was computed from another hash (first) from a large string would result in many collisions that will make it impossible to distinguish between them and the first hash of the password!
4. If possible use HMAC for added security! This will prevent sending the same generated hash from a paswrod in the tcp/udp packets... each time the pasword field in tha data packets will contain a different hash (HMAC'ed) but on server side will be decoded to the real hash. For this you will need to design a solution for sending the password needed for HMAC!
blodo at poczta dot fm
07-Jul-2005 05:05
You could always double hash the password, which might just be the easiest way to prevent a hash table crack which typically can only crack hashes that store about 8 characters in length (from what i know). For an additional amount of security you can always append a random salt, although you will need to store the salt to succesfully recover the password for checking. The code could look like this:
<?php

function createSalt ($charno = 5) {

  for (
$i = 0; $i < $charno; $i++;) {
  
  
$num = rand(48, 122); //This is roughly the ascii readable character range
  
$salt .= chr($num); //Convert the number to a real character
      
 
}
  
  return
$salt;
  
}

$randomsalt = createSalt();
$hash = md5(md5($string_to_hash.$randomsalt));

/* Now append this to your database, but remember to store the salt somewhere too, or else you wont be able to check the password later on */

?>

Easy as pie. Hope this helps.
Chris Ladd
06-Jun-2005 02:41
Using a hard coded salt will only prevent a dictionary attack on the raw database and only if the attacker hasn't gained access to the file the salt is stored in. The attacker could still use the php login site to run a dictionary attack, since the php would be adding the hard coded salt to the password, md5ing it, and checking if it matches. The only real way to prevent a dictionary attack is to not allow weak passwords. There is no shortcuts in real security.
oguh at gmx dot net
12-May-2005 02:06
Better use a random value for the salt and store it seperate in the database for every user.

So it is not possible to see if some users have the same password.
Jim Plush - jiminoc at gmail dot com
18-Mar-2005 05:45
Another handy trick is to use MD5 with a "salt". Which basically  means appending another static string to your $password variable to help prevent against dictionary attacks.

Example:
config.php - KEEP THIS OUTSIDE THE WEBROOT
define("PHP_SALT", "iLov3pHp5");
----------------------------------------------

and when you add your database query you would do:
// storing password hash
$query  = sprintf("INSERT INTO users(name,pwd) VALUES('%s','%s');",
           addslashes($username), md5($password.PHP_SALT));

This way if a user's password is "DOG" it can't be guessed easily because their password gets saved to the DB as the MD5 version of "DOGiLov3pHp5". Last time I checked, that wasn't in the dictionary :)