Package: ntp
Severity: wishlist
X-Debbugs-Cc: [email protected]
Control: subscribe -1
Hi Debian NTP maintainers--
Over on oss-security, it's been announced that some operators of IPv6
servers in the NTP pool who are using their position in the pool to
probe active IPv6 addresses.
One participant in the discussion proposed a mitigation technique where
NTP IPv6 clients could just allocate a dedicated IPv6 address that would
be otherwise firewalled and used only for NTP.
I think the mitigation proposal (included in full below) is actually
quite a nice idea, and something ideally suited for O/S distributions.
Is this something we could integrate into one of the debian packages
somehow?
You can see the message in context at:
http://openwall.com/lists/oss-security/2016/01/29/4
Regards,
--dkg
--- Begin Message ---
On 27 January 2016 at 14:43, Kurt Seifried <[email protected]> wrote:
> On Wed, Jan 27, 2016 at 4:24 AM, Luca BRUNO <[email protected]> wrote:
> > For oss-sec crowd: is there anything we can do to improve the situation
> > and avoid
> > similar cases in the future? Should crowd-sourced and fundamental services
> > like this
> > be encouraged to move to a stronger WoT?
>
> [...]
>
> Sadly we can't really rely on the IoT device makers to fix this, they have
> basically 0 incentive to prevent scanners from hitting their products
> (they're already sold, to late for the customer to make an informed
> decision).
I hope you'll forgive me making a modest proposal here, but it seems
to me that there might be an opportunity here for Linux distributions
that are upstream of IoT vendors to modify their default configuration
to address this.
My somewhat off-the-cuff suggestion would be to...
1. Add an *additional, secondary* IPv6 address to external interfaces that is:
-> a. generated in accordance with the IPv6 Privacy Extensions (i.e. RFC 4941)
-> b. firewalled by default against all traffic except NTP in either direction
2. Configure the NTP *client* to use this secondary address as the
source for outgoing NTP traffic, instead of the default address?
...thereby avoiding revealing the primary address of the host to
would-be scanners?
I realise that that is a rather drastic approach, and might be too
bold a change for Debian or RHEL, but perhaps in the case of
distributions like Raspbian which focus on IoT, it might be tenable?
Cheers,
Hazel
--- End Message ---
