mcrypt_encrypt()

(PHP 4 >= 4.0.2, PHP 5, PHP 7 = 1.0.0)

使用给定参数加密明文

This function has beenDEPRECATEDas of PHP 7.1.0. Relying on this functionis highly discouraged.

说明

加密数据并返回密文。

参数

MCRYPT_ciphername常量中的一个，或者是字符串值的算法名称。

加密密钥。如果密钥长度不是该算法所能够支持的有效长度，则函数将会发出警告并返回FALSE

使用给定的$cipher和$mode加密的数据。如果数据长度不是 n*分组大小，则在其后使用'0'补齐。

返回的密文长度可能比$data更大。

MCRYPT_MODE_modename常量中的一个，或以下字符串中的一个："ecb"，"cbc"，"cfb"，"ofb"，"nofb"和"stream"。

Used for the initialization in CBC, CFB, OFB modes, and in some algorithms in STREAM mode. If the provided IV size is not supported by the chaining mode or no IV was provided, but the chaining mode requires one, the function will emit a warning and return FALSE.

返回值

以字符串方式返回密文，或者在失败时返回FALSE。

更新日志

范例

mcrypt_encrypt()例程

以上例程会输出：

Key size: 32
ENJW8mS2KaJoNB5E5CoSAAu0xARgsR1bdzFWpEn+poYw45q+73az5kYi4j+0haevext1dGrcW8Qi59txfCBV8BBj3bzRP3dFCp3CPQSJ8eU=
This string was AES-256 / CBC / ZeroBytePadding encrypted.

参见

• mcrypt_decrypt()使用给定参数解密密文
• mcrypt_module_open()打开算法和模式对应的模块

If you're writing code to encrypt/encrypt data in 2015, you should use openssl_encrypt() and openssl_decrypt(). The underlying library (libmcrypt) has been abandoned since 2007, and performs far worse than OpenSSL (which leverages AES-NI on modern processors and is cache-timing safe).
Also, MCRYPT_RIJNDAEL_256 is not AES-256, it's a different variant of the Rijndael block cipher. If you want AES-256 in mcrypt, you have to use MCRYPT_RIJNDAEL_128 with a 32-byte key. OpenSSL makes it more obvious which mode you are using (i.e. 'aes-128-cbc' vs 'aes-256-ctr').
OpenSSL also uses PKCS7 padding with CBC mode rather than mcrypt's NULL byte padding. Thus, mcrypt is more likely to make your code vulnerable to padding oracle attacks than OpenSSL.
Finally, if you are not authenticating your ciphertexts (Encrypt Then MAC), you're doing it wrong.
Further reading:
https://paragonie.com/blog/2015/05/using-encryption-and-authentication-correctly
https://paragonie.com/blog/2015/05/if-you-re-typing-word-mcrypt-into-your-code-you-re-doing-it-wrong

Solving 3DES incompatibilities with .NET's TripleDESCryptoServiceProvider
mcrypt's 3DES only accepts 192 bit keys, but Microsoft's .NET and many other tools accept both 128 and 192 bit keys.
If your key is too short, mcrypt will 'helpfully' pad null characters onto the end, but .NET refuses to use a key where the last third is all null (this is a Bad Key). This prevents you from emulating mcrypt's "short key" behaviour in .NET.
How to reconcile this? A little DES theory is in order
3DES runs the DES algorithm three times, using each third of your 192 bit key as the 64 bit DES key
Encrypt Key1 -> Decrypt Key2 -> Encrypt Key3
and both .NET and PHP's mcrypt do this the same way.
The problem arises in short key mode on .NET, since 128 bits is only two 64 bit DES keys
The algorithm that they use then is:
Encrypt Key1 -> Decrypt Key2 -> Encrypt Key1
mcrypt does not have this mode of operation natively.
but before you go and start running DES three times yourself, here's a Quick Fix

And, like magic, it works.
There's one more caveat: Data padding
mcrypt always pads data will the null character
but .NET has two padding modes: "Zeros" and "PKCS7"
Zeros is identical to the mcrypt scheme, but PKCS7 is the default.
PKCS7 isn't much more complex, though:
instead of nulls, it appends the total number of padding bytes (which means, for 3DES, it can be a value from 0x01 to 0x07)
if your plaintext is "ABC", it will be padded into:
0x41 0x42 0x43 0x05 0x05 0x05 0x05 0x05
You can remove these from a decrypted string in PHP by counting the number of times that last character appears, and if it matches it's ordinal value, truncating the string by that many characters:

And to pad a string that you intend to decrypt with .NET, just add the chr() value of the number of padding bytes:

That's all there is to it.
Knowing this, you can encrypt, decrypt, and duplicate exactly any .NET 3DES behaviour in PHP.

The encryption has no authenticity check. It can be achieved with three methods, described in http://en.wikipedia.org/wiki/Authenticated_encryption#Approaches_to_Authenticated_Encryption
Encrypt-then-MAC (EtM), Encrypt-and-MAC (E&M), MAC-then-Encrypt (MtE).
The following is a suggestion for MtE: 

It is always better to use a standard encryption cipher's rather than to "roll your own", firstly the standard cipher has been tested by world class crypto-analysis's where as unless your a world class crypto-analysis (and if you are why are you even thinking of rolling your own?!?) you won't have the skills needed to even test it (for example if you just XOR each character with a key, it may look secure (the text will be different) but if you count the number of times a character is repeated you see whatever the letter "E" encrypts to occurs more often then the encrypted "Z" (assuming English language plain text)
Secondly, you may think that the hidden nature of your cipher makes it more secure, but the fact is that your cipher is likely *only* secure because it's secret, if someone what able to break-in to your site and steal your code (but not your key) they maybe able break you encrypted data, if someone broke in and found you where using Blowfish (for example) it wouldn't help them.

Note that the IV must be the same for mcrypt_encrypt and mcrypt_decrypt, otherwise you will have corrupted data after decryption.

I've noticed some people using a-z, A-Z and 0-9 for keys and stating things like "16 characters is a 128-bit key". This isn't true. Using only these characters, you will get at most 6 bits of entropy per chartacter:
log2(26 + 26 + 10) = 5.954196310386876
So you're actually only getting 95 bits of entropy in 16 characters, which is 0.0000000117% of the keyspace you would get if you were using the full range.
In order to get the full entropy from a key using just a-z, A-Z and 0-9 you should multiply your key length by 1.3333 to account for the 2 bits of lost entropy per byte.

In the other notes there are some misconceptions about crypto and the IV, especially for CBC mode.
The most important point: Encryption DOES NOT provide any proof of data integrity or authentication WHATSOEVER. If you need to be sure that the data is secret and not tampered with, you need to encrypt THEN use a keyed HMAC.
For CBC mode, the IV DOES NOT need to be secret. It can be sent along with the plaintext. It needs to be UNIQUE and RANDOM. So that every message is encrypted with a different IV.
The best way to generate an IV is to use mcrypt_create_iv().
Keys must be binary, not ASCII. To create a key from a password: 

I should mention that ECB mode ignores the IV, so it is misleading to show an example using both MCRYPT_MODE_ECB and an IV (the example in the manual shows the same thing). Also, it's important to know that ECB is useful for random data, but structured data should use a stronger mode like MCRYPT_MODE_CBC
Also, rtrim($decryptedtext, "\0") would be a better option to remove NULL padding than my lazy trim()...

Contrary to what is implied in the mcrypt_encrypt() manual page, as well as the info given regarding the CBC vs CFB modes, mcrypt_encrypt() works just fine for encrypting binary data as well.
 
A simple example verifies that the decrypted output is binary identical once cut to its original length: 

I was trying (and succeeded) to encrypt and decrypt in JAVA, pass it to php, and do it again,without corrupting data when I noticed something interesting. So, my code goes like this:
$data = 'one';
$key = '1234567890123456';
function encrypt($data, $key){
  return base64_encode(
  mcrypt_encrypt(
    MCRYPT_RIJNDAEL_128,
    $key,
    $data,
    MCRYPT_MODE_CBC,
    "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
  )
);
}
function decrypt($data, $key){
  $decode = base64_decode($data);
  return mcrypt_decrypt(
          MCRYPT_RIJNDAEL_128,
          $key,
          $decode,
          MCRYPT_MODE_CBC,
          "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
      );
  
  
}
$encrypted = encrypt($data, $key);
$decrypted= decrypt($encrypted, $key);
//In the beginning, I thought something's wrong, because I did '===' comparison between $decrypted and $data. It didn't work (but later started working, again, dont know why...) So, I dumped both:
var_dump($data);
var_dump($decrypted);
//Results:
string(16) "one" 
string(16) "one" 
//Clearly, the length of both is 3, not 16. Just wanted to let you know what could happen, and I really don't know if this is a bug... 
Thanks,
S.

Most of the user-written cipher examples here are badly broken, and there are a few cases where the manual says things that are outright incorrect, such as that it's "safe to transmit the initialization vector in plaintext" (this is incorrect: see Ciphers By Ritter, http://www.ciphersbyritter.com/GLOSSARY.HTM#IV , for details.)
mcrypt itself is perfectly safe, but correct and therefore safe usage is inobvious. It is important to use a cryptographic library correctly; a simple usage error, even when it produces results that can be unpacked at the other side, can render a strong algorithm completely useless.
The initialization vector must be permuted with a recoverable noise source (an arbitrary md5 hash is acceptable, since it's just a fake OTP and its origin contents are wholly unimportant.) 
Passwords should be remade with a salted one-way hash (md5 is again acceptable even though it's been damaged, since the only thing you could recover from a cracked md5 hash is the source data to generate the password, which is useless.) 
It's important to use a sane block mode (OFB is unsafe for almost all algorithms; never use it. Prefer CBC in all cases except where you need to deal with a degraded signal and cannot retransmit.)
A correct usage example is actually pretty long and needs a lot of explanation, so I developed a safe wrapper library which doesn't constrain usage and which comments itself very heavily. It's appropriate for use or for learning. Please see my blog for details on Stone PHP SafeCrypt:
http://blog.sc.tri-bit.com/archives/101

I always use this method to prevent a lot of errors
function encrypt( $string ) {
 $algorithm = 'rijndael-128'; // You can use any of the available
 $key = md5( "mypassword", true); // bynary raw 16 byte dimension.
 $iv_length = mcrypt_get_iv_size( $algorithm, MCRYPT_MODE_CBC );
 $iv = mcrypt_create_iv( $iv_length, MCRYPT_RAND );
 $encrypted = mcrypt_encrypt( $algorithm, $key, $string, MCRYPT_MODE_CBC, $iv );
 $result = base64_encode( $iv . $encrypted );
 return $result;
}
function decrypt( $string ) {
 $algorithm = 'rijndael-128';
 $key = md5( "mypassword", true );
 $iv_length = mcrypt_get_iv_size( $algorithm, MCRYPT_MODE_CBC );
 $string = base64_decode( $string );
 $iv = substr( $string, 0, $iv_length );
 $encrypted = substr( $string, $iv_length );
 $result = mcrypt_decrypt( $algoritmo, $key, $encrypted, MCRYPT_MODE_CBC, $iv );
 return $result;
}

Please note that the following part of the documentation is no longer true (after commit: http://git.php.net/?p=php-src.git;a=commit;h=a861a3a93d89a50ce58e1ab1abef1eb501f97483):
> key
> The key with which the data will be encrypted. If it's smaller than the required keysize, it is padded with '\0'. It is better not to use ASCII strings for keys.
That commit changed the behaviour to be strict and if the keysize is smaller than the required size a warning will be issued as follows:
Warning: mcrypt_encrypt(): Key of size 10 not supported by this algorithm. Only keys of size 16 supported in script.php on line 5
and the mcrypt_encode() will return failure.

I tried to implement the mcrypt with rijndael-128. For reference I took the code from example #1 and tried running that first, but on the decryption part came back with the error: "The IV parameter must be as long as the blocksize". After a while i figured out that the generated IV string will not have the same length every run, and is almost never the size of the result of mcrypt_get_iv_size. To work around that, before merging the IV and the encrypted text, I added null padding to match the IV size. When retrieving the IV, I then could use the IV size and rtrim null padding to get the matching IV back.
The altered parts from example #1 

鹏仔微信 15129739599 鹏仔QQ344225443 鹏仔前端 pjxi.com 共享博客 sharedbk.com