Re: Newbie: Key question

[Schuyler Stultz]

Ole Hansen wrote:

I now pass the 32 bytes (after the 5 bytes Record Layer Header) to my
decrypt function and I expected a result that at least had the handshake
protocol header as the first 4 bytes indicating the handshake type (20)
and length but it is not. Thats what I meant with things not making
sense. The result seems to be just as random as the input. I know a
hashvalue is part of the Finished message but does it not have a four
byte header as well? Maby thats the problem.
 

Yes, it will have the 4 byte Handshake Header, followed by 12 bytes of 
the verifyData, then the X byte HMAC, and finally, any padding (if 
needed.)  

So, for an example RSA 3DES EDE CBC SHA1 connection, you would have 40 
bytes that were encrypted:  
   4 bytes for the Handshake header which contains 0x14 for Finished, 
then 3 bytes of length (in this case, 0x0, 0x0, 0xc since it's just the 
length of the handshake's verify_data)
   12 bytes of verify_data
   20 bytes of the SHA1_HMAC
   3 bytes of padding
   1 byte of pad length
-------------------------------
   40 bytes of encrypted data.

I have used EVP_Decrypt*-functions to decrypt the Finished message. They
are part of openssl-0.9.7c. I think this should be ok but I'm not quite
sure how to use them. Not many examples from the openssl documentation.
 

Hmmmm  Be sure you're initializing the decrypt process with the server's 
write key, and the server's IVec.   Not the same keys you used to 
encrypt the client messages.  And one thing I also needed to do that 
wasn't immediately obvious to me was to set the 
EVP_CIPHER_CTX_set_padding(&cipherCtx, 0);  SSL/TLS padding is different 
than standard PKCS padding, so you should handle it instead of letting 
the EVP_Decrypt routines.  The Pad Length byte doesn't count in the 
padding.  So, in the above case, I would have 3 bytes of padding (set to 
3) then the pad byte is set to 3 as well.  Standard PKCS padding would 
have all of the padding bytes set to 4.  But, if you're doing a stream 
cipher (like RC4) you won't have padding...

(Now that I've said that, does the EVP_ routines handle SSL style padding?)

Anyway, hopefully one of these things will help point you in the right 
direction to find what your actual problem is...

Thanks, and a happy new year.

And to you as well!! :)

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
User Support Mailing List                    [EMAIL PROTECTED]
Automated List Manager                           [EMAIL PROTECTED]