[Meego-security-discussion] Backing up (was: MeeGo Security Goals)

[Rene Mayrhofer]

On 16.04.2011 00:01, Andy Ross wrote:
>>> Fine-grained network access limitation.  Facebook app can only
>>> talk to facebook, etc...
>> How *could* it work?
> Make access network go through a custom device and whitelist some IPs
> with an iptables plugin, make it go through a transparent proxy,
> etc...  If the app is written with this kind of limitation in mind and
> you trust the higher-level framework to set up the firewall, you can
> certainly limit, say, children of a "facebook" app process to
> connecting to IPs that reverse under "*.facebook.com" (or that result
> from forward lookups too, if you integrate with the local DNS
> resolver).
Or simply by using the uid/gid iptables match if we choose to adopt the 
Android model of "every app gets it own uid/gid on installation". But I 
agree that it's too early to discuss how to achieve certain security 
goals. We should first try to define which things we are trying to 
protect against, and only later on try to come up with potential 
implementations. It's possible that some security goals may be scratched 
at a later stage if the common decision is that it would be too complex 
for the base system or for application developers.

I propose to make the list of security goals a kind of "wishlist" at 
this point. Some may be mandatory, but some may get a tag such as "we 
should try to do this, but will not add a huge layer of complexity just 
for this one item, at least not in this version". Maybe in a Wiki, 
somewhere?

best regards.
Rene