RE: IBM OS/400 Multiple OSP-CERT Vulnerabilities

Agh.


> From: David Gibbs
> 
> I found this while stumbling around today ...
> 
> http://www.securityfocus.com/bid/14800/info
> 
>  IBM OS/400 osp-cert is susceptible to multiple vulnerabilities.

Whoever posted the message to SecurityFocus (and I think we all know who
that was, don't we?) was very careful in misstating and misrepresenting
the issues involved.


> The first identified vulnerability is a flaw in local Certificate
> Authority certificates. During the creation of these certificates,
X.509
> basic constraints are not added. Since the local Certificate Authority
> certificates are not properly identified as valid Certificate
> Authorities, the proper validation of these certificates is
impossible,
> as is the verification of certificate chains that are derived from it.

All this means is that anybody who is properly following RFC2459 (and
the subsequent RFC2632) will reject these certifications because they
don't have the mandatory basic constraints.  If you understand how basic
constraints work, they are used limit the role of a certification when
it is from a CA; if you don't see this extension, you don't accept the
cert as coming from a CA.  As client software tightens the rules, you
have to get your own processes up to speed.


> The next issue is a failure of the application to properly return
> renewed certificates once they have been made available. This issue is
> due to a failure of the application to properly notice altered
> certificate store files during operation.

This is simply an enhancement that fixes a long-standing issue in the
X.509 standard.


> There are also multiple unspecified ASN.1 parsing vulnerabilities. No
> further information about these issues are currently available.

This is indeed a vulnerability.  However, it is a vulnerability in the
X.509 standards, one that conceivably be used to cause a buffer overrun
in not only Solaris, but also virtually every Microsoft operating
system.  However, since buffer overruns simply don't happen on i5/OS,
it's not really an issue.  However, it could conceivably cause an
exception in the IKE server and thus cause a denial of service, so IBM
fixed it.


> These issues allow attackers to potentially subvert, bypass, or
possibly
> alter the security properties of the cryptographic software. The exact
> impact and possible attack scenarios are not known at this time.

This is just a lie.


For more information about the vulnerabilities:

http://www.ciac.org/ciac/bulletins/o-065.shtml
http://www.microsoft.com/technet/security/bulletin/MS04-011.mspx


Joe