public string GenerateToken(int length)
{
    using (RNGCryptoServiceProvider cryptRNG = new RNGCryptoServiceProvider())
    {
        byte[] tokenBuffer = new byte[length];
        cryptRNG.GetBytes(tokenBuffer);
        return Convert.ToBase64String(tokenBuffer);
    }
}

(你也可以让这个方法所在的类实现IDisposable，持有对RNGCryptoServiceProvider的引用，并正确地处理它，以避免重复实例化它。)

It's been noted that as this returns a base-64 string, the output length is always a multiple of 4, with the extra space using = as a padding character. The length parameter specifies the length of the byte buffer, not the output string (and is therefore perhaps not the best name for that parameter, now I think about it). This controls how many bytes of entropy the password will have. However, because base-64 uses a 4-character block to encode each 3 bytes of input, if you ask for a length that's not a multiple of 3, there will be some extra "space", and it'll use = to fill the extra.

If you don't like using base-64 strings for any reason, you can replace the Convert.ToBase64String() call with either a conversion to regular string, or with any of the Encoding methods; eg. Encoding.UTF8.GetString(tokenBuffer) - just make sure you pick a character set that can represent the full range of values coming out of the RNG, and that produces characters that are compatible with wherever you're sending or storing this. Using Unicode, for example, tends to give a lot of Chinese characters. Using base-64 guarantees a widely-compatible set of characters, and the characteristics of such a string shouldn't make it any less secure as long as you use a decent hashing algorithm.

由于Random是不安全的，RNGCryptoServiceProvider是过时的，我最终这样做:

  // possible characters that password can have
  private const string passChars =
       "ABCDEFGHIJKLMNOPQRSTUVWXYZ" +
       "abcdefghijklmnopqrstuvwxyz" +
       "0123456789" +
       "!@#$%.-_"
       ;

    public static string GetRandomPassword(int length)
    {
        char[] p = new char[length];
        for (int i = 0; i < length; i++)
            p[i] = passChars[RandomNumberGenerator.GetInt32(0, passChars.Length)];
        return new string(p);
    }

public string GenerateToken(int length)
{
    using (RNGCryptoServiceProvider cryptRNG = new RNGCryptoServiceProvider())
    {
        byte[] tokenBuffer = new byte[length];
        cryptRNG.GetBytes(tokenBuffer);
        return Convert.ToBase64String(tokenBuffer);
    }
}

(你也可以让这个方法所在的类实现IDisposable，持有对RNGCryptoServiceProvider的引用，并正确地处理它，以避免重复实例化它。)

It's been noted that as this returns a base-64 string, the output length is always a multiple of 4, with the extra space using = as a padding character. The length parameter specifies the length of the byte buffer, not the output string (and is therefore perhaps not the best name for that parameter, now I think about it). This controls how many bytes of entropy the password will have. However, because base-64 uses a 4-character block to encode each 3 bytes of input, if you ask for a length that's not a multiple of 3, there will be some extra "space", and it'll use = to fill the extra.

If you don't like using base-64 strings for any reason, you can replace the Convert.ToBase64String() call with either a conversion to regular string, or with any of the Encoding methods; eg. Encoding.UTF8.GetString(tokenBuffer) - just make sure you pick a character set that can represent the full range of values coming out of the RNG, and that produces characters that are compatible with wherever you're sending or storing this. Using Unicode, for example, tends to give a lot of Chinese characters. Using base-64 guarantees a widely-compatible set of characters, and the characteristics of such a string shouldn't make it any less secure as long as you use a decent hashing algorithm.

我不喜欢Membership.GeneratePassword()创建的密码，因为它们太丑了，有太多特殊字符。

这段代码生成一个10位不太难看的密码。

string password = Guid.NewGuid().ToString("N").ToLower()
                      .Replace("1", "").Replace("o", "").Replace("0","")
                      .Substring(0,10);

当然，我可以使用一个Regex来做所有的替换，但这是更具可读性和可维护性的IMO。

public string CreatePassword(int length)
{
        const string valid = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890";
        StringBuilder res = new StringBuilder();
        Random rnd = new Random();
        while (0 < length--)
        {
            res.Append(valid[rnd.Next(valid.Length)]);
        }
        return res.ToString();
}

这有一个很好的好处，可以从可用的字符列表中选择生成的密码(例如，仅数字，仅大写或仅小写等)。

在接受的答案中添加了一些补充代码。它改进了仅使用随机的答案，并允许一些密码选项。我也喜欢KeePass回答中的一些选项，但不想在我的解决方案中包含可执行文件。

private string RandomPassword(int length, bool includeCharacters, bool includeNumbers, bool includeUppercase, bool includeNonAlphaNumericCharacters, bool includeLookAlikes)
{
    if (length < 8 || length > 128) throw new ArgumentOutOfRangeException("length");
    if (!includeCharacters && !includeNumbers && !includeNonAlphaNumericCharacters) throw new ArgumentException("RandomPassword-Key arguments all false, no values would be returned");

    string pw = "";
    do
    {
        pw += System.Web.Security.Membership.GeneratePassword(128, 25);
        pw = RemoveCharacters(pw, includeCharacters, includeNumbers, includeUppercase, includeNonAlphaNumericCharacters, includeLookAlikes);
    } while (pw.Length < length);

    return pw.Substring(0, length);
}

private string RemoveCharacters(string passwordString, bool includeCharacters, bool includeNumbers, bool includeUppercase, bool includeNonAlphaNumericCharacters, bool includeLookAlikes)
{
    if (!includeCharacters)
    {
        var remove = new string[] { "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z" };
        foreach (string r in remove)
        {
            passwordString = passwordString.Replace(r, string.Empty);
            passwordString = passwordString.Replace(r.ToUpper(), string.Empty);
        }
    }

    if (!includeNumbers)
    {
        var remove = new string[] { "0", "1", "2", "3", "4", "5", "6", "7", "8", "9" };
        foreach (string r in remove)
            passwordString = passwordString.Replace(r, string.Empty);
    }

    if (!includeUppercase)
        passwordString = passwordString.ToLower();

    if (!includeNonAlphaNumericCharacters)
    {
        var remove = new string[] { "!", "@", "#", "$", "%", "^", "&", "*", "(", ")", "-", "_", "+", "=", "{", "}", "[", "]", "|", "\\", ":", ";", "<", ">", "/", "?", "." };
        foreach (string r in remove)
            passwordString = passwordString.Replace(r, string.Empty);
    }

    if (!includeLookAlikes)
    {
        var remove = new string[] { "(", ")", "0", "O", "o", "1", "i", "I", "l", "|", "!", ":", ";" };
        foreach (string r in remove)
            passwordString = passwordString.Replace(r, string.Empty);
    }

    return passwordString;
}

这是我搜索生成随机密码时的第一个链接，以下内容超出了当前问题的范围，但可能很重要，需要考虑。

假设System.Web.Security.Membership.GeneratePassword是加密安全的，至少有20%的字符是非字母数字。 不确定在这种情况下删除字符和追加字符串是否被认为是良好的实践，并提供足够的熵。 可能需要考虑以某种方式使用SecureString实现内存中的安全密码存储。