Re: [OpenAFS] What you need to know about Windows 10

2015-07-29 Thread Antoine Verheijen
On Jul 29, 2015, at 9:10 AM, Jeffrey Altman  
wrote:

> On 7/29/2015 3:12 AM, Antoine Verheijen wrote:
>> Putting my security hat on: certified drivers does not provide ANY
>> additional degree of security whatsoever. It merely states that the
>> certifier has blessed it using whatever criteria they use (in many
>> cases, simply financial payment).
>> 
>> What guarantee(s) is the certifier prepared to live up to via their
>> certification? If none, why is it required?
> 
> Certification provides quality control.  Microsoft's signing of the
> kernel drivers does not involve any payment.  Microsoft is willing to
> sign any drivers that have passed the required quality control checks
> which include test suites, static analysis, and feature/capability lists.

I'll accept this point at face value, in particular as I have no direct
experience with Microsoft in this regard. Furthermore, I realize in
hindsight that this not the venue to discuss an issue of this sort as it
does not relate in any meaningful way to AFS, the real subject of this
mailing list, and I should never have made my initial comments in this
discussion list. I apologize for having done so.

> The only additional security benefit of Microsoft signing the drivers as
> opposed to permitting vendors to use issued cross signing certificates
> is that a vendor cannot longer be hacked and have their signing key be
> used without their knowledge to sign unapproved binaries without a paper
> trail.

This is a totally valid point, one which I had not considered, and which
most certainly does provide increased security (albeit perhaps not of the
sort I had in mind), clearly contradicting my initial assertion. :-)

> Jeffrey Altman

Once again, apologies for the inappropriate content. I'll try to be more
considerate. :-)

Bye for now.


Antoine VerheijenEmail: [email protected]
.Phone: (780) 
462-9696___
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] What you need to know about Windows 10

2015-07-29 Thread Jeffrey Altman
On 7/29/2015 3:12 AM, Antoine Verheijen wrote:
> Putting my security hat on: certified drivers does not provide ANY
> additional degree of security whatsoever. It merely states that the
> certifier has blessed it using whatever criteria they use (in many
> cases, simply financial payment).
> 
> What guarantee(s) is the certifier prepared to live up to via their
> certification? If none, why is it required?

Certification provides quality control.  Microsoft's signing of the
kernel drivers does not involve any payment.  Microsoft is willing to
sign any drivers that have passed the required quality control checks
which include test suites, static analysis, and feature/capability lists.

The only additional security benefit of Microsoft signing the drivers as
opposed to permitting vendors to use issued cross signing certificates
is that a vendor cannot longer be hacked and have their signing key be
used without their knowledge to sign unapproved binaries without a paper
trail.

Jeffrey Altman




smime.p7s
Description: S/MIME Cryptographic Signature


Re: [OpenAFS] What you need to know about Windows 10

2015-07-29 Thread Daria Brashear
On Wed, Jul 29, 2015 at 9:56 AM, Brandon Allbery 
wrote:

> On Wed, 2015-07-29 at 01:12 -0600, Antoine Verheijen wrote:
> > What guarantee(s) is the certifier prepared to live up to via their
> > certification? If none, why is it required?
>
> It is a point... Apple is apparently willing to let anyone request a
> kext signing certificate,


I gather the requests are vetted even if the kexts are not directly so


-- 
D


Re: [OpenAFS] What you need to know about Windows 10

2015-07-29 Thread Brandon Allbery
On Wed, 2015-07-29 at 01:12 -0600, Antoine Verheijen wrote:
> What guarantee(s) is the certifier prepared to live up to via their
> certification? If none, why is it required?

It is a point... Apple is apparently willing to let anyone request a
kext signing certificate, whereas Microsoft requires you to submit the
extension to them for signing and presumably validate it in some way.

-- 
brandon s allbery kf8nh   sine nomine associates
[email protected]  [email protected]
unix openafs kerberos infrastructure xmonadhttp://sinenomine.net


Re: [OpenAFS] What you need to know about Windows 10

2015-07-29 Thread Antoine Verheijen
Putting my security hat on: certified drivers does not provide ANY
additional degree of security whatsoever. It merely states that the
certifier has blessed it using whatever criteria they use (in many
cases, simply financial payment).

What guarantee(s) is the certifier prepared to live up to via their
certification? If none, why is it required?

Just my thoughts. Thanks for listening. :-)


Antoine VerheijenEmail: [email protected]
.Phone: (780) 462-9696

On Jul 29, 2015, at 12:48 AM, Gary Buhrmaster  wrote:

> On Wed, Jul 29, 2015 at 12:28 AM, Jeffrey Altman
>  wrote:
> 
>> Tomorrow(*)
> 
> Thanks for the update/reminder.  And thanks for your
> willingness to build "one last time" for Windows 10.
> It really is "above and beyond" what anyone has any
> right to expect.
> 
> Personally, I have no idea if Windows 10 will be
> everything MS wants it to be(**), but (putting my security
> hat on) moving to certified drivers is the right way
> forward, regardless of how it impacts some projects
> (and those projects need to "step up their game").
> 
> Thanks.
> 
> Gary(***)
> 
> 
> (*) The right statement to many on Wednesday really should be:
>"and then, and then, do the smart thing, let someone else try first"
> 
> (**) "Its tough to make predictions, especially about the future"
> 
> (***) Can't find a quote to steal over my sig.
> ___
> OpenAFS-info mailing list
> [email protected]
> https://lists.openafs.org/mailman/listinfo/openafs-info

___
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] What you need to know about Windows 10

2015-07-28 Thread Gary Buhrmaster
On Wed, Jul 29, 2015 at 12:28 AM, Jeffrey Altman
 wrote:

> Tomorrow(*)

Thanks for the update/reminder.  And thanks for your
willingness to build "one last time" for Windows 10.
It really is "above and beyond" what anyone has any
right to expect.

Personally, I have no idea if Windows 10 will be
everything MS wants it to be(**), but (putting my security
hat on) moving to certified drivers is the right way
forward, regardless of how it impacts some projects
(and those projects need to "step up their game").

Thanks.

Gary(***)


(*) The right statement to many on Wednesday really should be:
"and then, and then, do the smart thing, let someone else try first"

(**) "Its tough to make predictions, especially about the future"

(***) Can't find a quote to steal over my sig.
___
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info


[OpenAFS] What you need to know about Windows 10

2015-07-28 Thread Jeffrey Altman
Tomorrow is Wednesday July 29 and the day that Microsoft is going to
release Windows 10 to approximately 5 million users that have either
been a part of the Windows Insider program or pre-registered for a free
upgrade.  Windows 10 will be made available to volume license customers
on August 1.  Some vendors such as Dell and Lenovo will begin shipping
pre-loaded systems tomorrow and stores such as Best Buy have been
encouraged to upgrade current stock to Windows 10 before they let the
machines out the door.  The USB Flash Disk images will be shipping on
August 30.

There will not be a build of OpenAFS 1.7 targeted at Windows 10
available on the release day.   I am hoping to produce what will be my
last "OpenAFS" branded client with support for Windows 10 by the start
of the AFS and Kerberos Best Practices Workshop on August 17th.  After
that I will only be releasing AuriStor branded clients and I will
explain why at the end of this letter.

First, what do I know about the existing 1.7.32 build and Windows 10.

1. The 1.7.32 build does work (for the most part) on Windows 10 but

1a. the installation will be damaged during an upgrade from Windows 7 or
Windows 8.1 to Windows 10.  In particular, the network provider
registration will be lost.  End users should be encouraged to run
"Repair" on the OpenAFS components after the installation is complete.

1b. there are some changes to the method by which the afs redirector is
accessed that can under some circumstance result in a BSOD.

2. The infamous Explorer Shell caching bug that resulted in reports that
there are 0 bytes free when copying files to \\AFS has been fixed in
Windows 10.

3. As a result of the Explorer Shell bug being fixed the AFS redirector
needs to be modified to undo the hack that disabled the reporting of
read only volume state.

4. There is another known bug in shell32.dll that has not been fixed
that can result in a deadlock if a UNC path such as
\\afs\share-does-not-exist\ is entered into the explorer shell or into a
file open/save dialog box.   I have a workaround to implement in OpenAFS
but it is not ready.

5. There are known bugs in the AFS redirector or service that can

5a. prevent failover to alternative .readonly volume sites

5b. result in access to the wrong file object if two or more objects
exist with names that differ only by case in the same directory

6. The Netbios interface that the afsd_service relies for the SMB server
interface has been removed in Windows 10.   As a result the AFS SMB
interface must be permanently disabled when running on Windows 10.

7. Windows 10 supports UNC hardening for secure access to roaming
profiles and network based executables and configuration files.
Microsoft best practice states that UNC hardening should be turned on.
UNC hardening protects against man in the middle attacks that can result
in execution of untrusted code or the loading of untrusted user registry
hives by the system.  OpenAFS does not support UNC hardening and it must
remain disabled.

By the workshop I plan to have an OpenAFS based installer to distribute.
 This installer will not be signed by Microsoft but by the older
cross-signing certificate method.

By the workshop I also hope to demo the first AuriStor based client
which will:

1. support UNC hardening

2. support IPv6 connectivity

3. include a new kernel driver to process ICMP messages for faster
   failover and detection of IPv6 Path MTU sizes.

4. be compiled with Visual Studio 2015

5. be signed by Microsoft

This client will be the client that I am going to submit to Microsoft
for certification testing.  It is my hope that certification approvals
will be issued by October 16th which is expected to be the day that
production quality previews of Server 2016 will be released.  As I have
mentioned previously, only drivers that were signed by Microsoft and
include a certification attributed in the signature can be loaded on
forthcoming Windows Server releases.

Support for Server Nano will not be completed by October.  I am hoping
that can be completed by Spring 2016.

Jeffrey Altman



smime.p7s
Description: S/MIME Cryptographic Signature