All of the network's control points are on public non-firewalled IPs.
This is the worst security. It was done this way for the sake of
simplicity. Our netops volunteers had to get up to speed with
unfamiliar concepts like routing, funky netmasks, dynamic routing
protocols, policy routing, VRRP, firewalls, MTUs, MSS control, IPsec,
etc. We reaped the rewards of KISS from broader volunteer engagement,
but lately we've been paying too heavy of a price for the awful security
this simplicity creates. In the most recent breach we've lost important
source code that will now need to be re-created. We escaped total
disaster by the thinnest of margins, as one critical hypervisor just
happened to be patched to 1 version higher than exploitable. This
simplicity is not a good tradeoff anymore, so the time has come to
introduce more complexity to the network to protect all control points.
This is not a simple problem, since there are many fragility vs security
tradeoffs, as well as complexity cost concerns. If you have experience
or thoughts around this area, and can commit to a few weeks of design
and implementation work on this project, please indicate your interest.
We'll assemble a small working group in the next few days and start
discussions. I expect the working format will involve some virtual
meetings, since email is not high bandwidth enough to hash out
everything quickly.
Here's hoping we don't make it worse,
--Bart
_______________________________________________
PSDR mailing list