1996-12-26 - RE: [NOT NOISE] Microsoft Crypto Service Provider API

Header Data

From: Blake Coverett <blake@bcdev.com>
To: “cypherpunks@toad.com>
Message Hash: 51b952bc8390bbc210d9e07667bb12a3c8f988b22e57f6ecc4fa38c1c110afe5
Message ID: <01BBF2B9.DBC67AB0@bcdev.com>
Reply To: N/A
UTC Datetime: 1996-12-26 04:18:04 UTC
Raw Date: Wed, 25 Dec 1996 20:18:04 -0800 (PST)

Raw message

From: Blake Coverett <blake@bcdev.com>
Date: Wed, 25 Dec 1996 20:18:04 -0800 (PST)
To: "cypherpunks@toad.com>
Subject: RE: [NOT NOISE] Microsoft Crypto Service Provider API
Message-ID: <01BBF2B9.DBC67AB0@bcdev.com>
MIME-Version: 1.0
Content-Type: text/plain


> Not quite.  The API comes with a program SIGN.EXE that will create a
> "debugging signature" for your CSP, and a new ADVAPI32.DLL, described as
> a "Modified advapi32.dll to load providers that are signed with
> sign.exe."  So the patch point is a bit more accessable than inside the
> kernel.  Maybe the "Modified advapi32.dll" should find its way offshore?

Even better than exporting the hacked advapi32.dll, compare the it 
with the original one.  I'd bet good money that the only difference is
the contents of the RC_DATA/#102 resource attached to the image.
(It's useful to note that the advapi32.dll from versions of NT before
CryptoAPI doesn't have any RC_DATA resources.)

And to think MS was good enough to provide an UpdateResource 
API that I haven't yet had a good reason to use.

> Interestingly enough, CSP signatures are held in the registry instead of
> the binary, necessitating some install procedure for a given CSP.  Not
> to start rumors, but NT 4.0 does use threads to watch some registry
> entries that control the version (workstation/server).  Not much of a
> stretch to imagine a thread that tracks (reports?) changes to

Nope, a little experimentation shows you can change those entries
while the system is running to your hearts contents.  Try temporarily
renaming the signature key of the base provider.

regards,
-Blake






Thread