Re: Conficker scanning with nmap

[David Fifield <david () bamsoftware com>]

On Wed, Apr 01, 2009 at 11:26:25AM -0500, Ron wrote:
> David Fifield wrote:
> > I can reproduce this with OpenSSL 0.9.8e and the attached sample script.
> > The error only happens for me when encrypting an empty string, not
> > decrypting one.
> >
> > I think you're right about where the problem occurs. It could happen
> > when the server sends a zero-length challenge. The challenge comes from
> > line 615 of smbauth.lua. It looks like it could be caused by a truncated
> > packet.
>
> I can probably default the challenge to "AAAAAAAA" or something.. a  
> blank challenge shouldn't happen anyway, it's likely the server would  
> ignore the answer either way.
>
> I've added a check to my smb.lua class (haven't committed it yet) that  
> automatically changes a blank server challenge to 'AAAAAAAA'.
>
> Do we maybe want to add this check deeper, though? Like, in  
> openssl.encrypt(), do a check on the data + the version and return an  
> error (or return known bad data) if we end up in the situation?

I think we should just avoid calling EVP_EncryptUpdate when data_len == 0.
That's what the patch that Lionel found does, just bails out if the
length is <= 0.

http://cvs.openssl.org/filediff?f=openssl/crypto/evp/evp_enc.c&v1=1.45&v2=1.46

However we probably need to make the other calls (EVP_EncryptInit_ex,
EVP_CIPHER_CTX_set_padding, EVP_EncryptFinal_ex), as they might be
necessary for some ciphers. Encrypting the empty string may still
require writing an initialization vector or something like that.

David Fifield

_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://SecLists.Org