linux-kernel - Re: [PATCH 3/3] x86: kernel base offset ASLR

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

Date:	Thu, 4 Apr 2013 13:47:50 -0700
From:	Eric Northup <digitaleric@...gle.com>
To:	"H. Peter Anvin" <hpa@...or.com>
Cc:	Kees Cook <keescook@...omium.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	kernel-hardening@...ts.openwall.com,
	Thomas Gleixner <tglx@...utronix.de>,
	Ingo Molnar <mingo@...hat.com>,
	"the arch/x86 maintainers" <x86@...nel.org>,
	Jarkko Sakkinen <jarkko.sakkinen@...el.com>,
	Matthew Garrett <mjg@...hat.com>,
	Matt Fleming <matt.fleming@...el.com>,
	Dan Rosenberg <drosenberg@...curity.com>,
	Julien Tinnes <jln@...gle.com>, Will Drewry <wad@...omium.org>
Subject: Re: [PATCH 3/3] x86: kernel base offset ASLR

On Thu, Apr 4, 2013 at 1:21 PM, H. Peter Anvin <hpa@...or.com> wrote:
> I have to admit to being somewhat skeptical toward KASLR with only 8
> bits of randomness.

I agree that 8 bits is pretty low and more would be better.  However,
even 8 bits provides a < 1% chance that any particular guess will be
correct.  Combined with kernel crash monitoring, this amount of ASLR
means that brute-force attacks can't occur undetectably, even if they
can eventually be successful.  Having a signal that indicates an
attack-in-progress is a pretty big leg up from not.  Of course,
infoleaks would render this whole discussion moot, but I'm replying to
the "only 8 bits" part here.

> There are at least two potential ways of
> dramatically increasing the available randomness:
>
> 1. actually compose the kernel of multiple independently relocatable
>    pieces (maybe chunk it on 2M boundaries or something.)

Without increasing the entropy bits, does this actually increase the #
of tries necessary for an attacker to guess correctly?  It
dramatically increases the number of possible configurations of kernel
address space, but for any given piece there are only 256 possible
locations.

> 2. compile the kernel as one of the memory models which can be executed
>    anywhere in the 64-bit address space.  The cost of this would have
>    to be quantified, of course.

I attempted to do this, but was limited by my knowledge of the
toolchain.  I would welcome help or suggestions!
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/