linux-kernel - Re: [PATCH v2 0/4] Checkpoint/Restore: Show in proc IDs of objects that can be shared between tasks

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 for Android: free password hash cracker in your pocket

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

Date:	Sat, 19 Nov 2011 09:30:07 +0400
From:	Cyrill Gorcunov <gorcunov@...il.com>
To:	Kyle Moffett <kyle@...fetthome.net>
Cc:	Tejun Heo <tj@...nel.org>, Pekka Enberg <penberg@...nel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Pavel Emelyanov <xemul@...allels.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Glauber Costa <glommer@...allels.com>,
	Andi Kleen <andi@...stfloor.org>,
	Matt Helsley <matthltc@...ibm.com>,
	Eric Dumazet <eric.dumazet@...il.com>,
	Vasiliy Kulikov <segoon@...nwall.com>
Subject: Re: [PATCH v2 0/4] Checkpoint/Restore: Show in proc IDs of objects
 that can be shared between tasks

On Fri, Nov 18, 2011 at 08:09:12PM -0500, Kyle Moffett wrote:
...
> >
> > The new version is using different poison for different types of
> > objects.
> 
> Even still, if you use a one-time pad (IE: XOR with a random data
> value) to obscure more than exactly 1 object total, ever, all of its
> security properties are null and void.
>

True. It's not one-time pads there.

> 
> >> If you actually want to be able to compare uniqueness without exposing
> >> anything vulnerable to various kinds of guessing, you should generate
> >> a random 64-bit value for each class of object and then use a proper
> >> cryptographic hash function on it:
> >>   crypto_hash(concat(object_ptr, random_value))
> >>
> >> Even given the best possible practical attacks against SHA1 or MD5
> >> today both still provides more than enough security to prevent someone
> >> from guessing "object_ptr" in less than an absurd amount of time.
> >
> > So, per-type poison + crypto hash, it is then.
> 
> Yes.  I haven't thought through whether or not you would ever care
> about a system giving out the same number for two different kinds of
> object.  The only possible vulnerability I can think of would be if
> the kernel had a use-after-free bug... You could allocate and free a
> bunch of the vulnerable objects and use this data-structure-ID system
> to find an allocated data-structure of a different type which matches
> up with one of the used-after-freed ones.  Then in theory you could
> compromise something, I suppose.
> 
> Sort of an off-the-wall scenario, I will admit.
> 
> The per-type random value is certainly a safe bet and should have zero
> actual impact on performance.  Good luck!
> 

Thanks for all comments Kyle! Of course address allocation specifics with
simple xor wont give us enough obscurity here. If we stick with root-only
approach then we don't need this scheme at all but could expose plain
addresses. I'm waiting for Pavel's comment on such approach.

	Cyrill
--
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/