Hello,

On Sat, Feb 27, 2021 at 6:19 PM Richard Guy Briggs <rgb@redhat.com> wrote:
On 2021-02-26 15:21, Andreas Hasenack wrote:
Issue ghak124 (https://github.com/linux-audit/audit-kernel/issues/124)
introduced auditing for nftables modifications.  It turns out it was far
too verbose but may have listed these actions for the iptables-nft
variant.  That is about to be trimmed but should still catch any
changes for nftables.

What parameters do you wish to have logged?  At a quick look, I'm
guessing table doesn't make sense since a set could be used by any
registered table?  But the set name would, followed by protocol family,
number of items changed, and the operation name?

I'm not sure if there are regulatory requirements about what has to be logged in this case, but yeah, what caught my eye is that a firewall rule can effectively be changed by just changing the ipset it references, and that change didn't trigger a NETFILTER_CFG audit message. This is with iptables, not nftables. I don't know if it's handled differently with nftables.

How much life does iptables have to it?  Given that this command can

You mean for how long will people still be using iptables? I'm not sure, but I personally bet in a few more years.

 
change the configuration of iptables (and ipv6tables, ebtables,...) it
would seem this this should be logged.

That was my thinking, but I thought about a log of its own, not part of iptables. To be honest I haven't checked yet what changes in NETFILTER_CFG with nftables, if anything. I know custom rules catching setsockopt won't catch nftables changes, but that's about it.