Client/server security

I think, I think (I hope :-)... If Informix had implemented password checking on the set role statment we could possibly have solved the security issues that have reasantly been discussed in a very easy and elegant way. The statement would be: set role rolename identified by password We could then store the password somewhere in an encrypted form. It could be in a database table every user had select rights only to. The application would then have to decrypt the password, plug it into the above statement and execute it. We could write encryption and decryption algoritms so that it whould not be easy for a user to decrypt the password. In this way the user wouldn't be able to set the role outside of an application that have the built in functionallity to do it. I can see three possible holes in this: 1. If you use the current I-Net or some primitive ODBC driver the above set role statement would be sent in the clear over the net. It would be relatively easy to see it. With Openlink ODBC drivers there is an encryption option that could solve this. A malisious user could still run an ODBC spy on his own machine and see the above statement. 2. The encryption/decryption algorithm could be difficult or impossible to write in such a way that someone couldn't do the same thing. Additionally if it was done in a DLL on MS Windows it wouldn't be too hard to use the functions in the DLL in some other program. Even without documentation for the DLL. 3. Someone with access to the real encryption/decryption algorithms could use them to gain access to a database that he should not have had. These are still only problems for those who require a fairly high level of security. If you do there may be solutions using the Universal Server. A datablade, spl or possibly Java code in the server could be used to make the whole procedure fully secure - possibly. Does anyone see any problems with the above? (It's so easy to forget some stupid little thing that opens glaring wholes in sceems like this.) Problem 3 above may be the most difficult to solve. Any concrete ideas? Could one make it secure via a public key algorithm or in some other way - even to those with access to the algorithms? Whether these problems can be solved fully or not, I do at least believe we could get another level of security in this way. Please respond even if you think it would work. If it works I will talk to Informix and get them to implement such a password. (I am an optimistic person you see :-) I will also urge everybody else to do the same - call or mail Informix and say you need this. But *please* wait until we have gotten some responses so we know that I haven't suggested something stupid! Nils.Myklebust@idg.no NM Data AS, P.O.Box 9090 Gronland, N-0133 Oslo, Norway My opinions are those of my company The Informix FAQ is at http://www.iiug.org