[HELP] Password encryption

[Qonquer]

The password encryption is basically the same as before with some added extras. The server now sends a extra 32 bit cypher before login starts, the base key has changed, there is a character by character substitution before encryption, and a further encryption based on the ascii sum of the pz user name.

I did get quite far into working it out before other commitments took me away. If I get back on it in the near future I'll probably post here with it.

[unknownone]

Yeah, it's still RC5. The key to seed it now is based on that int received from the server. It's used to create some 16 random bytes (srand(), rand() from msvcrt), and the 16 byte sequence is what seeds the RC5.

In addition to the RC5 though, there's another layer of encryption, which is seeded from the sum of all the username's character's ASCII values, and that seed then creates a 256 int16 key, which is then sorted by most sig byte only. The cryptography then works based on key indexes and converting scan codes to virtual key characters (ie, MapVirtualKey());

.Net implementation of that part here anyway. Reflect it if you need to port to another language. It has no external dependencies.

Also an RC5 implementation for those who were lazy and hardcoded the old key rather than implementing the keygen routine.

Credit to Ultimation & Myself.

[Neo~]

Originally Posted by unknownone

Actually, they previously used RC5, and more recently, began using their own algorithm (which still uses part of the RC5, with modifications).

Why are you such a nerd?

[~Falcon]

Originally Posted by Neo~

Why are you such a nerd?

He knows his stuff, which is very fortunate for us, as he most likely makes up for all the retards here.

[PeTe Ninja]

Originally Posted by Neo~

Why are you such a nerd?

just because hes smarter than you doesn't make him a nerd. just a highly advanced human being. either that or your just retarded.

[.Guru]

Originally Posted by Zion~

It depends on the version. From 4267 to 5017 TQ used the blowfish(cipher) algorithm and for their latest patch a modified version of RSA is used (as far as I know).

5165 updated lotf version also uses blowfish

[Qonquer]

Originally Posted by unknownone

Yeah, it's still RC5. The key to seed it now is based on that int received from the server. It's used to create some 16 random bytes (srand(), rand() from msvcrt), and the 16 byte sequence is what seeds the RC5.

In addition to the RC5 though, there's another layer of encryption, which is seeded from the sum of all the username's character's ASCII values, and that seed then creates a 256 int16 key, which is then sorted by most sig byte only. The cryptography then works based on key indexes and converting scan codes to virtual key characters (ie, MapVirtualKey());

.Net implementation of that part here anyway. Reflect it if you need to port to another language. It has no external dependencies.

Also an RC5 implementation for those who were lazy and hardcoded the old key rather than implementing the keygen routine.

Credit to Ultimation & Myself.

Nice one, lol that saved me some time in ollydbg. I knew you guys woulda sussed it.

Originally Posted by Neo~

Why are you such a nerd?

I guarantee he's smarter than you. Probably by several times. Contrary to normal belief, someone smarter than you (which may be a great proportion of the population) is not a nerd.

[~Yuki~]

i myselfe would like to know how i can decrypt the Pass be4 writing it...

I need code peaces please

[Qonquer]

It's all in unknownone's dll he attached earlier in this post. Just use a reflector to open the dll and you can see the source in any language you prefer. You should really thank him for this not me mang.

[~Yuki~]

fixed thanked you both

[unknownone]

The only part missing from the DLL is the creation of the RC5 key. It's just an int sent from the server to the client on connect in an 0x423 packet. For a basic template of how you'd implement it..

    using Liberate.Cryptography;
    using msvcrt;

    class Connection
    {
        private readonly int passCryptoSeed;
        //...

        public Connection(Socket socket)
        {
            this.passCryptoSeed = (new Random()).Next();
            var seedPacket = new PasswordCryptographySeedPacket()
            {
                Seed = this.passCryptoSeed
            };
            this.SendPacket(seedPacket);
            //...
        }

        public void OnLoginPacket(LoginPacket loginPacket)
        {
            msvcrt.srand(this.passCryptoSeed);
            var rc5Key = new byte[0x10];
            for (int i = 0; i < 0x10; i++) 
                rc5Key[i] = (byte)msvcrt.rand();
            var password = Encoding.ASCII.GetString(
                                (new ConquerPasswordCryptpographer(loginPacket.Username)).Decrypt(
                                    (new RC5(rc5Key)).Decrypt(loginPacket.Password)));
            //...
        }
        //...
    }

[Qonquer]

Not dome much .NET stuff myself, just a little worried about the seeding of the PRNG like that. Are the variables used in srand() and rand() private to that particular thread or are they global for the entire application? If the latter is true, it could cause problems to a whole server application to have the seed constantly reset to a relatively small integer everytime someone logs on. A multi threaded application could also call rand() right in the middle of the key generation causing a failed login. Of course, if the routines are pure to the thread it isn't an issue.

[XtremeX-CO]

Originally Posted by Qonquer

Not dome much .NET stuff myself, just a little worried about the seeding of the PRNG like that. Are the variables used in srand() and rand() private to that particular thread or are they global for the entire application? If the latter is true, it could cause problems to a whole server application to have the seed constantly reset to a relatively small integer everytime someone logs on. A multi threaded application could also call rand() right in the middle of the key generation causing a failed login. Of course, if the routines are pure to the thread it isn't an issue.

public class rand
{

    static int _seed;
    int seed;

    public rand(int seed)
    {
        this.seed = seed;
    }

    public final int _next()
    {
        seed *= 0x343fd;
        seed += 0x269Ec3;
        return (seed >> 0x10) & 0x7FFF;
    }

    public static void seed(int seed)
    {
        _seed = seed;
    }

    public static final int next()
    {
        _seed *= 0x343fd;
        _seed += 0x269Ec3;
        return (_seed >> 0x10) & 0x7FFF;
    }

    public static final int next(int max)
    {
        return (next() % (max + 1));
    }

    public static final int next(int min, int max)
    {
        return (next() % (int) (((max) + 1) - (min))) + (min);
    }
}

could just do
rand rand = new rand(seed)
rand._next();


or static mode
rand.seed(seed)
rand.next()


That's in Java, basically a copy of what msvcrt does because Java's random sucks



edit: Also, I'm having some problems with my encryptions still, it decrypts correctly about half the letters (and all upper case). I think the mapping may be wrong, not sure (using sparkie's source).

For example:
Encrypted (by client): teest
Decrypted (by proxy): ( S( (contains 2 spaces after ( )

Encrypted (by client): aaabbb
Decrypted (by proxy): ���BBB