As I do not think that the D-bus transport will be used in a final
product where the socket transport will provide more the performances
which are expected, we need to define a security model which is not
realted to D-Bus but rather generic.

Something in the still of the Application manifest translated to the
CAN Zones would likely be a viable model.

Thoughts ?

2014-07-09 14:03 GMT+02:00 Patrick Ohly <[email protected]>:
> On Mon, 2014-07-07 at 14:46 -0700, Rees, Kevron wrote:
>> i've listed some use cases that define WHAT needs to be secure in
>> Automotive Message Broker (AMB).  The next question is "HOW"...
>>
>> https://wiki.tizen.org/wiki/AMB_Security
>>
>> The security policy engine needs to know the user and zone of the
>> requesting process id.  Can cynara be used like this?
>
> Cynara at the moment only knows about privileges; AMB concepts like
> individual properties and zones seem to be a poor match for that. But
> I'll let you discuss that with John and the Cynara developers. Please
> report back to the list with a summary.
>
> My question is something else: AMB uses a publish/subscribe model for
> property changes, and one transport for that is D-Bus. See
> org.freedesktop.DBus.Properties.PropertiesChanged in
> https://wiki.tizen.org/wiki/AMB_DBus_Plugin.
>
> If we stick to the original plan (services call cynara or do their own
> privilege checking), then you cannot use D-Bus signals with unset
> destination and rely on dbus-daemon to route the signal to interested
> recipients, because your service won't know who is subscribed via a
> watch filter (*) and the dbus-daemon doesn't know who of the subscribers
> are allowed to receive the signal.
>
> (*) The Wiki page mentions that AMB tracks if someone is interested in a
> property change, but that's not what dbus-daemon uses.
>
> You could send out one PropertiesChanged signal per client calling
> Manager.Find, addressed specifically to that client, but that is less
> efficient.
>
> Have you thought about deprecating the
> org.freedesktop.DBus.Properties.PropertiesChanged API in favor of some
> other transport? Not sure whether that is an option, considering that
> D-Bus is the IPC mechanism used in GENIVI.
>
> How often do such signals get emitted? D-Bus may be a poor match also
> for performance reasons.
>
> If you need to keep the current API and want to avoid sending one signal
> per recipient, then dbus-daemon needs be enhanced such that it can check
> whether a certain subscriber is allowed to receive a certain signal. In
> user-space dbus-daemon, this could take the object path into account. It
> cannot be based on the actual property, because that is a detail that
> dbus-daemon doesn't know about.
>
> In KDBus it'll be even worse; there the object path is part of the
> opaque payload and thus not available for routing decisions.
>
> --
> Best Regards, Patrick Ohly
>
> The content of this message is my personal opinion only and although
> I am an employee of Intel, the statements I make here in no way
> represent Intel's position on the issue, nor am I authorized to speak
> on behalf of Intel on this matter.
>
>
>
> _______________________________________________
> IVI mailing list
> [email protected]
> https://lists.tizen.org/listinfo/ivi



-- 
Dominig ar Foll
Senior Software Architect
Intel Open Source Technology Centre
_______________________________________________
IVI mailing list
[email protected]
https://lists.tizen.org/listinfo/ivi

Reply via email to