Has anyone written anything about the problem of using OpenSSL in an untrusted
environment? If there is nothing written I would consider studying it and
sharing the results.
Here's the problem. Suppose you want to communicate securely with an
UNTRUSTED party who can authenticate themselves to you.
A concrete example: Alice National Bank wants to transfer some funds to
customer Bob's digital wallet. The bank gives Bob a wallet program (linked
with OpenSSL) that talks to the bank server by SSL. So the transaction is
authenticated and secure from third parties who would steal money or find out
the contents of Bob's transaction. The problem here is to prevent Bob from
hacking the wallet by adjusting the wallet program's environment in order to
get more money than he's supposed to. (Forget about signed digital money for
this question).
In the case of OpenSSL, its environment looks vulnerable. Some obvious holes
for a hostile user that is *NOT* going as far as decompiling the wallet:
- the environment variable $SSL_CIPHER could be used to specify a weaker
cipher than the bank wants;
- $RANDFILE or /dev/random (or egd) could be replaced with a predictable
source;
- whatever else the CONF facility lets the user specify;
- whatever else someone who studies the code can figure out.
OpenSSL is obviously designed for flexibility and configuration at both ends
of a connection. Questions:
- Has anyone cataloged all these untrusted-site-holes?
- Has anyone added a compile-time option to block them all?
- Is anyone else interested in untrusted site usage?
Thanks for any thoughts!
Respectfully,
--
Mitchell Perilstein
[EMAIL PROTECTED]
www.enetis.net/~mitch
+1 (605) 574-2367
______________________________________________________________________
OpenSSL Project http://www.openssl.org
User Support Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
