Guy
Guy

Reputation: 320

C# & OpenSSL: Differing Output for AES Encryption

I'm attempting to write a C# method that will spit out the same encrypted string (Base64) as the openssl binary does, but am having a heck of a time getting things to match up.

Lots of terminal-output and C# to follow... :P

We're going to use the very exciting example of encrypting the string "a" with the password "123".

First is openssl when I provide my static salt and a password (this is how the command will be ideally run, and is what I want my C# output to match to):

dev@magoo ~# echo -n a | openssl enc -aes-128-cbc -S cc77e2a591358a1c -pass pass:123 -a -p
salt=CC77E2A591358A1C
key=7B2AD689138A44AD32297BBAAA5B0EEE
iv =EC4F0416B2E9A9B2FEEF2E66FF982159
U2FsdGVkX1/Md+KlkTWKHOtnt1ftHSKyWiz6GxkBVck=

Second is openssl when I provide my static salt, and the key and iv that are derived from that salt (C+P'd from the output of the first command) but no password (since even the docs say that would be a not-good idea):

dev@magoo ~# echo -n a | openssl enc -aes-128-cbc -S cc77e2a591358a1c -K 7b2ad689138a44ad32297bbaaa5b0eee -iv ec4f0416b2e9a9b2feef2e66ff982159 -a -p
salt=E85778B7FFFFFFFF
key=7B2AD689138A44AD32297BBAAA5B0EEE
iv =EC4F0416B2E9A9B2FEEF2E66FF982159
62e3V+0dIrJaLPobGQFVyQ==

This strikes me as odd. Adding the key and iv values from the "debug" output (-p param) in the first command to the same salt, I somehow get a different salt! (CC77E2A591358A1C vs E85778B7FFFFFFFF [the 4 bytes of 0xff here seem interesting maybe]).

Third is the output of my application:

dev@magoo ~# mono aestest.exe "a" "123"
==> INPUT     : a
==> SECRET    : 123
==> SALT      : cc77e2a591358a1c
==> KEY       : 7b2ad689138a44ad32297bbaaa5b0eee
==> IV        : ec4f0416b2e9a9b2feef2e66ff982159
==> ENCRYPTED : 62e3V+0dIrJaLPobGQFVyQ==

So, the C# matches the output of the openssl command when I manually specified the key and IV myself (that then somehow generated a different salt), but that seems wrong somehow. In my mind the C# application's output should match to the first set of OpenSSL's output, shouldn't it?

C# code:

public static string EncryptString(string plainText, string password)
{
  byte[] salt = Encryption.GetStaticSalt();
  byte[] key, iv;

  Encryption.DeriveKeyAndIV(password, salt, out key, out iv);

  var amAes = new AesManaged();
  amAes.Mode = CipherMode.CBC;
  amAes.KeySize = 128;
  amAes.BlockSize = 128;
  amAes.Key = key;
  amAes.IV = iv;

  var icTransformer = amAes.CreateEncryptor();
  var msTemp = new MemoryStream();

  var csEncrypt = new CryptoStream(msTemp, icTransformer, CryptoStreamMode.Write);
  var sw = new StreamWriter(csEncrypt);
  sw.Write(plainText);
  sw.Close();
  sw.Dispose();

  csEncrypt.Clear();
  csEncrypt.Dispose();

  byte[] bResult = msTemp.ToArray();
  string sResult = Convert.ToBase64String(bResult);

  if (System.Diagnostics.Debugger.IsAttached)
  {
    string debugDetails = "";
    debugDetails += "==> INPUT     : " + plainText + Environment.NewLine;
    debugDetails += "==> SECRET    : " + password + Environment.NewLine;
    debugDetails += "==> SALT      : " + Encryption.ByteArrayToHexString(salt) + Environment.NewLine;
    debugDetails += "==> KEY       : " + Encryption.ByteArrayToHexString(amAes.Key) + " (" + amAes.KeySize.ToString() + ")" + Environment.NewLine;
    debugDetails += "==> IV        : " + Encryption.ByteArrayToHexString(amAes.IV) + Environment.NewLine;
    debugDetails += "==> ENCRYPTED : " + sResult;
    Console.WriteLine(debugDetails);
  }

  return sResult;
}

private static string ByteArrayToHexString(byte[] bytes)
{
  StringBuilder sbHex = new StringBuilder();

  foreach (byte b in bytes)
    sbHex.AppendFormat("{0:x2}", b);

  return sbHex.ToString();
}

public static byte[] GetStaticSalt()
{
  // Just random bytes.
  return new byte[]
  {
    0xcc,
    0x77,
    0xe2,
    0xa5,
    0x91,
    0x35,
    0x8a,
    0x1c
  };
}

// largely hijacked from http://stackoverflow.com/a/8011654/97423
public static void DeriveKeyAndIV(string password, byte[] bSalt, out byte[] bKey, out byte[] bIV)
{
  int keyLen = 16;
  int ivLen = 16;

  byte[] bPassword = Encoding.UTF8.GetBytes(password);

  using (var md5Gen = MD5.Create())
  {
    List<byte> lstHashes = new List<byte>(keyLen + ivLen);
    byte[] currHash = new byte[0];

    int preHashLength = bPassword.Length + bSalt.Length;
    byte[] preHash = new byte[preHashLength];

    Buffer.BlockCopy(bPassword, 0, preHash, 0, bPassword.Length);
    Buffer.BlockCopy(bSalt, 0, preHash, bPassword.Length, bSalt.Length);

    currHash = md5Gen.ComputeHash(preHash);
    lstHashes.AddRange(currHash);

    while (lstHashes.Count < (keyLen + ivLen))
    {
      preHashLength = currHash.Length + password.Length + bSalt.Length;
      preHash = new byte[preHashLength];

      Buffer.BlockCopy(currHash, 0, preHash, 0, currHash.Length);
      Buffer.BlockCopy(bPassword, 0, preHash, currHash.Length, password.Length);
      Buffer.BlockCopy(bSalt, 0, preHash, currHash.Length + password.Length, bSalt.Length);

      currHash = md5Gen.ComputeHash(preHash);
      lstHashes.AddRange(currHash);
    }

    bKey = new byte[keyLen];
    bIV = new byte[ivLen];

    lstHashes.CopyTo(0, bKey, 0, keyLen);
    lstHashes.CopyTo(keyLen, bIV, 0, ivLen);
  }
}

Am I missing something really obvious here, or is this something more subtle? I've looked around SO and have seen a lot about C#, OpenSSL and AES, but nothing about this specific issue... so, halp? ;)

Upvotes: 5

Views: 2850

Answers (1)

Maarten Bodewes
Maarten Bodewes

Reputation: 93978

If you specify a key and IV directly, then the salt does not even get into play. The salt is needed to convert the pass phrase into a key using a key derivation function (the proprietary EVP_BytesToKey in the case of OpenSSL). Hence you probably get inconsequential output for the salt.

Now the first output of OpenSSL in 1) contains a header (check the ASCII values :), the salt and then the cipher text, this is the base 64 string in hexadecimals:

53616C7465645F5F CC77E2A591358A1C EB67B757ED1D22B25A2CFA1B190155C9

Your application 3) and the openssl command in 2) both output

EB67B757ED1D22B25A2CFA1B190155C9

so every little thing seems to be alright.

Upvotes: 1

Related Questions