18
votes

I'm having a problem with C# encrypting and decrypting using RSA. I have developed a web service that will be sent sensitive financial information and transactions. What I would like to be able to do is on the client side, Encrypt the certain fields using the clients RSA Private key, once it has reached my service it will decrypt with the clients public key.

At the moment I keep getting a "The data to be decrypted exceeds the maximum for this modulus of 128 bytes." exception. I have not dealt much with C# RSA cryptography so any help would be greatly appreciated.

This is the method i am using to generate the keys

private void buttonGenerate_Click(object sender, EventArgs e)
{
    string secretKey = RandomString(12, true);

    CspParameters param = new CspParameters();
    param.Flags = CspProviderFlags.UseMachineKeyStore;

    SecureString secureString = new SecureString();
    byte[] stringBytes = Encoding.ASCII.GetBytes(secretKey);
    for (int i = 0; i < stringBytes.Length; i++)
    {
        secureString.AppendChar((char)stringBytes[i]);
    }
    secureString.MakeReadOnly();
    param.KeyPassword = secureString;

    RSACryptoServiceProvider rsaProvider = new RSACryptoServiceProvider(param);
    rsaProvider = (RSACryptoServiceProvider)RSACryptoServiceProvider.Create();
    rsaProvider.KeySize = 1024;


    string publicKey = rsaProvider.ToXmlString(false);
    string privateKey = rsaProvider.ToXmlString(true);

    Repository.RSA_XML_PRIVATE_KEY = privateKey;
    Repository.RSA_XML_PUBLIC_KEY = publicKey;

    textBoxRsaPrivate.Text = Repository.RSA_XML_PRIVATE_KEY;
    textBoxRsaPublic.Text = Repository.RSA_XML_PUBLIC_KEY;

    MessageBox.Show("Please note, when generating keys you must sign on to the gateway\n" +
        " to exhange keys otherwise transactions will fail", "Key Exchange", MessageBoxButtons.OK, MessageBoxIcon.Information);

}

Once i have generated the keys, i send the public key to the web service which stores it as an XML file.

Now i decided to test this so here is my method to encrypt a string

public static string RsaEncrypt(string dataToEncrypt)
{
    string rsaPrivate = RSA_XML_PRIVATE_KEY;
    CspParameters csp = new CspParameters();
    csp.Flags = CspProviderFlags.UseMachineKeyStore;

    RSACryptoServiceProvider provider = new RSACryptoServiceProvider(csp);

    provider.FromXmlString(rsaPrivate);

    ASCIIEncoding enc = new ASCIIEncoding();
    int numOfChars = enc.GetByteCount(dataToEncrypt);
    byte[] tempArray = enc.GetBytes(dataToEncrypt);
    byte[] result = provider.Encrypt(tempArray, true);
    string resultString = Convert.ToBase64String(result);
    Console.WriteLine("Encrypted : " + resultString);
    return resultString;
}

I do get what seems to be an encrypted value. In the test crypto web method that i created, i then take this encrypted data, try and decrypt the data using the clients public key and send this back in the clear. But this is where the exception is thrown. Here is my method responsible for this.

public string DecryptRSA(string data, string merchantId)
{
    string clearData = null;
    try
    {
        CspParameters param = new CspParameters();
        param.Flags = CspProviderFlags.UseMachineKeyStore;
        RSACryptoServiceProvider rsaProvider = new RSACryptoServiceProvider(param);

        string merchantRsaPublic = GetXmlRsaKey(merchantId);
        rsaProvider.FromXmlString(merchantRsaPublic);
        byte[] asciiString = Encoding.ASCII.GetBytes(data);

        byte[] decryptedData = rsaProvider.Decrypt(asciiString, false);

        clearData = Convert.ToString(decryptedData);
    }
    catch (CryptographicException ex)
    {
        Log.Error("A cryptographic error occured trying to decrypt a value for " + merchantId, ex);

    }
    return clearData;
}

If anyone could help me that would be awesome, as i have said i have not done much with C# RSA encryption/decryption.

5
Does this thing about RSA encryption being cracked recently impact you? engadget.com/2010/03/09/…Nick Miller
RSA is not cracked. Only its implementation, in an actual system, is subject to side-channel attacks: the implementation may leak information on the private key through power consumption, timing, and other physical measures which are not part of the abstract model of a computer. This has been known for quite some time (timing attacks were described in 1997, I think). Such news is no news, and no real "crack" either.Thomas Pornin

5 Answers

36
votes

Allow me a bit of terminology. There is asymmetric encryption and there is digital signature.

  • Asymmetric encryption is about keeping confidentiality. Some sensitive data is transformed into something unreadable, save for the entity who knows the decryption key. The decryption key is necessarily the private key: if the decryption key is the public key, then everybody can decrypt the data (the public key is, well, public) and there is no confidentiality anymore. In asymmetric encryption, one encrypts with the public key and decrypts with the corresponding private key.

  • Digital signatures are meant to prove integrity. Someone computes a kind of keyed checksum over the data, in such a way that the link between the checksum and the data can be verified later. This is a "signature" only because the power to compute that checksum requires knowledge of something which is not public -- in plain words, signing uses the private key. Verification, however, should be doable by anybody, and thus use the public key.

A fair bit of confusion is implied by the fact that "the" RSA algorithm is actually a mathematical operation which can be declined into both an asymmetric encryption system, and a digital signature system. The confusion is further enhanced by the RSA standard, aka PKCS#1, which implicitly relies on how RSA digital signatures were first described, i.e. as a "reversed encryption" ("the signer encrypts the data with his private key"). Which leads to things like RSA signatures called "sha1WithRSAEncryption". This is quite unfortunate.

Therefore, you must first decide whether you want confidentiality or signatures. For confidentiality, for data sent from clients to the server, the server shall own a private key, and the clients use the server public key to encrypt the data. For signatures, each client shall have his own private key and use it to sign the data, and the server verifies the signatures. From your description I cannot tell what you are really after, thanks to the confusion I allude to above.

Also, there is something called authentication which may look like digital signatures, but is weaker. The point of signatures is than anybody can verify the signature. In particular, the signature can be shown to a judge and thus serve as legal weapon against the signer (the signature is legally binding -- at least if you do it right, and in the current state of regulations over electronic signatures, this is not easy). In most situations you only need something weaker and simpler, in which the server is convinced that it talks to the right client, but cannot afterwards convince anybody else that this client was really there. Any web site with user passwords is using such authentication.

With that being said...

  • RSA asymmetric encryption covers only short messages. For a 1024-bit RSA key (i.e. a key where the most important part, the "RSA modulus", is a big number with a value between 2^1023 and 2^1024, and encrypted messages will be of length 128 bytes), the maximum size of an encrypted message is 117 bytes (that's the actual source of your error message). When we want to send longer messages, we use an hybrid system, in which we only encrypt a small bunch of random bits (say 128 bits) and use that bunch as a key for a symmetric encryption system (e.g. AES) which can process much longer messages (and much faster, too).

  • RSA signatures, similarly, can be computed only on short messages, hence the PKCS#1 standard mandates that a signature is actually computed over a hash value. The hash value is the output of a specific hash function, which is computed over the message to sign. The hash function has a fixed-sized output (e.g. 256 bits for SHA-256) but accepts input messages of (almost) arbitrary length. Hash functions are public (there is no key in them) and, for proper security, must have some special properties. SHA-256 is, right now, not a bad choice. SHA-1 (a predecessor of SHA-256) has been proven to have some weaknesses and should be avoided. MD5 has (a kind-of uncle of SHA-1) has bigger weaknesses and shall not be used.

  • Proper use of asymmetric encryption, especially in an hybrid scheme, and digital signatures, is trickier than what the text above may suggest. It is awfully easy to get it wrong at some point, invisibly, i.e. the code will appear to work but will leak data useful for an attacker. The right way to use asymmetric encryption or digital signatures is to rely on existing, well-thought protocols. A protocol is an assembly of cryptographic elements into a coherent system, where leaks are taken care of. The prime example is TLS, also known as SSL. It is a protocol which ensures confidential data transmission, with integrity and authentication (possibly mutual authentication). The HTTPS protocol is a mixture of HTTP and SSL. The bright side is that HTTPS has existing implementations, notably in C#. The code which is easiest to implement and debug is the code which has already been implemented and debugged. So use HTTPS and you will live longer and happier.

15
votes

I understand why you are asking the question. The problem is that RSA is not used like a typical block cypher (like AES or 3DES) that encrypts 8 bytes at a time, all day long. RSA is a math operation that returns the remainder of a division (the modulus). Back in grade school, when you learned long division, remember that the remainder can never be greater than the divisor:if you are dividing 20 by 7, your remainder is 6. No matter what integer you divide by 7, the remainder cannot be greater than six.

RSA math is the same way. For example, if you are using a 1024-bit RSA public key, the remainder can never be greater than 2^1024, which is only 128 bytes. So you can only encrypt 128 bytes at a time with this key. (That's one reason we measure the size of RSA keys by the number of bits.)

Technically you could use this RSA key in a loop to encrypt 128 byte chunks of your data at a time. In reality, we almost never do this because RSA math is BIG and SLOW. Instead, we use what is called "two-phase" encryption. We use RSA to encrypt only a short "session key", and then use that session key in a fast symmetric-keyed block cypher (like AES) to encrypt the actual data.

The whole protocol is:

  1. Obtain the RSA public key of your destination. This is often delivered embedded in a certificate; if it is, be sure to validate the certificate to make sure the key is genuine. Let's say the RSA key is 2048 bits long.
  2. Generate a cryptographically strong pseudo-random number to use as a key for the block cypher (you need 256 bits as the key for AES-256, for example.) Note that 256 < 2048, the max that RSA-2048 can encrypt at once. We call this random number the "session key".
  3. Encrypt the session key using the RSA 2048-bit public key. It will give you 2048 bits of encrypted session key. Note that this operation is very slow.
  4. Encrypt all the secret data using AES-256, using the session key. Note that this is much faster than step 3.
  5. Bundle the public key ID from the certificate, the RSA encrypted session key, and the AES encrypted data together. I'd also tag it with a format identifier and version number, so you know what format it is in and how to decrypt it.
  6. Send the bundle to the destination.

  7. At the destination you use the format identifier and version to take apart the bundle.

  8. Retrieve the private key whose identity is in the public key ID field.
  9. Use this private key in RSA to decrypt the session key.
  10. Use the session key in AES to decrypt the data.

If you are going to do this, you should know that it is exactly what the CMS (PKCS#7) format is for. I would encourage you to learn about the standard and adopt it, rather than trying to invent your own format. Microsoft's CSP supports it, so it should be easy.

If you don't follow a standard you will have to make your own decisions about things like "what format should the AES key bits be in in the RSA encryption process?" More likely, you would almost certainly make security mistakes, weakening your system. Besides, you will find that tools such as the CSP will be very difficult to work with if you don't follow a standard.

3
votes

In DecryptRSA, is "data" base 64 encoded? If it is, you have to undo that first.

Honestly I think you shouldn't implement that routine yourself to protect "sensitive financial information", unless you have a lot of experience with cryptography. There are too many ways to make errors. Better use some ready solution - maybe SSL and certificates, or just PGP or GnuPG?

3
votes

RSA is primarily used to validate secure hashes of data - rather than encrypting the data itself. So, given a large blob of data, you might use SHA512 to create a hash of that data, then use RSA to sign that hash.

You'll want to use a symmetric encryption algorithm for large blocks of data - something like AES or 3DES.

Managing secure transactions isn't easy and really ought to be left to those guys that spend all day and night thinking about it. If you're exposing the service as over the web, just use SSL which already encrypts and secures your data.

2
votes

First decide what you are trying to protect against. If you "encrypt" something using the private key, anyone can "decrypt" it with the public key, since the public key is - well - public.

If you actually want to sign it, you should (as Paul Alexander explains) sign a hash with the private key which can then be verified on the server.

To encrypt data using RSA you should first generate a random symmetric key (f.x. AES), encrypt the key using a public key and encrypt the data using the symmetric key. You can then transmit the encrypted key together with the encrypted data to the holder of the private key, who can then first decrypt the encrypted key with the private key and then decrypt the data with the symmetric key.

You might also consider using SSL, but remember to carefully consider the authentication. You will probably need client authentication and have to decide which certificates to trust (you should not just blindly accept any certificate issued by Verisign).