phc-discussions - Re: [PHC] A little nit which bothers me...

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]

Message-ID: <56422.1396438686@critter.freebsd.dk>
Date: Wed, 02 Apr 2014 11:38:06 +0000
From: "Poul-Henning Kamp" <phk@....freebsd.dk>
To: discussions@...sword-hashing.net, Solar Designer <solar@...nwall.com>
cc: Brandon Enright <bmenrigh@...ndonenright.net>
Subject: Re: [PHC] A little nit which bothers me...

In message <20140402082306.GA24793@...nwall.com>, Solar Designer writes:

>Data-dependent branching has not been used as much yet, [...]

I never saw the point really.

First, the construct needs to be symmetic in order to not leak
too much timing information:

	if (CONDITION(entropy)) {
		entropy = FUNC1(entropy);
		discard = FUNC2(entropy);
	} else {
		discard = FUNC1(entropy);
		entropy = FUNC2(entropy);
	}

But there are so much junk between the CPU and the memory these
days that I still expect timing attacks can demask the condition bit.

The way to solve that is to only condition data-dependent
branching on "dead-end bits" which don't lead back to our entropy:

	if (CONDITION(HASH(entropy || "Lorem ipsum dolor sit..."))) {
		entropy = FUNC1(entropy);
		discard = FUNC2(entropy);
	} else {
		discard = FUNC1(entropy);
		entropy = FUNC2(entropy);
	}

We cannot lower our standards and allow timing attacks against
"dead-end" bits:

	if (CONDITION(HASH(entropy || "Lorem ipsum dolor sit..."))) {
		entropy = FUNC1(entropy);
	} else {
		entropy = FUNC2(entropy);
	}

That would give a 50% discount on a brute-force attack with timing
information.

In brute-force attack without timing information, the "discard"
bits could obviously be optimised out, that's also no good, so lets
not discard those bits anyway:

	if (CONDITION(HASH(entropy || "Lorem ipsum dolor sit..."))) {
		entropy[first_half] = FUNC1(entropy);
		entropy[second_half] = FUNC2(entropy);
	} else {
		entropy[second_half] = FUNC1(entropy);
		entropy[first_half] = FUNC2(entropy);
	}

And then it follows trivially that this is the same as:

	i = CONDITION(HASH(entropy || "Lorem ipsum dolor sit..."))
	entropy = FUNC12(entropy);
	if (i)
		swap_first_and_second_half(entropy)

Ohh well...

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@...eBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.