在与JAVA对接时,他们用的开源的Md5Crypt.md5Crypt加的密,以下为源码。
麻烦那位大神写个C#版本的。
public class Md5Crypt { /** The Identifier of the Apache variant. */ static final String APR1_PREFIX = "$apr1$"; /** The number of bytes of the final hash. */ private static final int BLOCKSIZE = 16; /** The Identifier of this crypt() variant. */ static final String MD5_PREFIX = "$1$"; /** The number of rounds of the big loop. */ private static final int ROUNDS = 1000; public static String md5Crypt(final byte[] keyBytes, final String salt, final String prefix) { final int keyLen = keyBytes.length; // Extract the real salt from the given string which can be a complete hash string. String saltString; if (salt == null) { saltString = B64.getRandomSalt(8); } else { final Pattern p = Pattern.compile("^" + prefix.replace("$", "\\$") + "([\\.\\/a-zA-Z0-9]{1,8}).*"); final Matcher m = p.matcher(salt); if (!m.find()) { throw new IllegalArgumentException("Invalid salt value: " + salt); } saltString = m.group(1); } final byte[] saltBytes = saltString.getBytes(Charsets.UTF_8); final MessageDigest ctx = DigestUtils.getMd5Digest(); /* * The password first, since that is what is most unknown */ ctx.update(keyBytes); /* * Then our magic string */ ctx.update(prefix.getBytes(Charsets.UTF_8)); /* * Then the raw salt */ ctx.update(saltBytes); /* * Then just as many characters of the MD5(pw,salt,pw) */ MessageDigest ctx1 = DigestUtils.getMd5Digest(); ctx1.update(keyBytes); ctx1.update(saltBytes); ctx1.update(keyBytes); byte[] finalb = ctx1.digest(); int ii = keyLen; while (ii > 0) { ctx.update(finalb, 0, ii > 16 ? 16 : ii); ii -= 16; } /* * Don't leave anything around in vm they could use. */ Arrays.fill(finalb, (byte) 0); /* * Then something really weird... */ ii = keyLen; final int j = 0; while (ii > 0) { if ((ii & 1) == 1) { ctx.update(finalb[j]); } else { ctx.update(keyBytes[j]); } ii >>= 1; } /* * Now make the output string */ final StringBuilder passwd = new StringBuilder(prefix + saltString + "$"); finalb = ctx.digest(); /* * and now, just to make sure things don't run too fast On a 60 Mhz Pentium this takes 34 msec, so you would * need 30 seconds to build a 1000 entry dictionary... */ for (int i = 0; i < ROUNDS; i++) { ctx1 = DigestUtils.getMd5Digest(); if ((i & 1) != 0) { ctx1.update(keyBytes); } else { ctx1.update(finalb, 0, BLOCKSIZE); } if (i % 3 != 0) { ctx1.update(saltBytes); } if (i % 7 != 0) { ctx1.update(keyBytes); } if ((i & 1) != 0) { ctx1.update(finalb, 0, BLOCKSIZE); } else { ctx1.update(keyBytes); } finalb = ctx1.digest(); } // The following was nearly identical to the Sha2Crypt code. // Again, the buflen is not really needed. // int buflen = MD5_PREFIX.length() - 1 + salt_string.length() + 1 + BLOCKSIZE + 1; B64.b64from24bit(finalb[0], finalb[6], finalb[12], 4, passwd); B64.b64from24bit(finalb[1], finalb[7], finalb[13], 4, passwd); B64.b64from24bit(finalb[2], finalb[8], finalb[14], 4, passwd); B64.b64from24bit(finalb[3], finalb[9], finalb[15], 4, passwd); B64.b64from24bit(finalb[4], finalb[10], finalb[5], 4, passwd); B64.b64from24bit((byte) 0, (byte) 0, finalb[11], 2, passwd); /* * Don't leave anything around in vm they could use. */ // Is there a better way to do this with the JVM? ctx.reset(); ctx1.reset(); Arrays.fill(keyBytes, (byte) 0); Arrays.fill(saltBytes, (byte) 0); Arrays.fill(finalb, (byte) 0); return passwd.toString(); } }