full-disclosure - firewall security bug?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <200401051359.i05DxQLG022823@caligula.anu.edu.au>
From: avalon at caligula.anu.edu.au (Darren Reed)
Subject: firewall security bug?

So reading through the spam that comes my way, I find this an
interesting read:

[...]
> >>8) States can be bound to the specific interface that created them or
> >>to  a group of interfaces for example:
> >>
> >>- pass all keep state (if-bound)
> >>- pass all keep state (group-bound)
> >>- pass all keep state (floating)
> >>
> >
> >Could you elaborate on this change and its uses a little? Especially the 
> >"group-bound" thing?
> >
> Ok. "floating" is the default, and is what PF has been doing all the time.
> That mean that if you've a rule like:
> 
>     pass in on fxp0 keep state
> 
> Once the state is created, PF will match that state with packets having
> the same characteristics (source/dest IP, same port for UDP/TCP, ...)
> coming in/out *any* interface. So, if a state is created for IKE traffic
> coming in on one interface, then any IKE packet with spoofed IP will be
> able to enter the firewall from any other interface, regardless of PF
> rules on that interface.
[...]

Is it just me or does that actually say there is a really nasty
security bug in PF's state tracking code making it impossible to
keep out spoof'd packets that match a state session, up until
this feature being introduced ?  And even then, having introduced
a mechanism to address it, it continues to default to off...maybe
the risk or danger posed by this is considered to be "low" but
even still, it seems to me to be a pretty fundamental kind of flaw
that should not even exist in the first place, IMHO.

Doing a good ol' RTFS of 3.3 seems to confirm this but my analysis
just might be biased ;)

Darren