News for hackers. Stuff that matters.
Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @11:09AM (#7298372)
This is absurd and unconscionable. There is no excuse whatsoever for not having a 10.9 patch ready to go at the same time as iOS. There is especially no excuse for still not having a 10.9 patch five days later.

I would love to see a detailed postmortem about exactly how this bug happened in the first place and why it's taking so long to fix it on the Mac side. Unfortunately, given how secretive the company is, I'm sure we'll never have more than speculation.

Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @11:24AM (#7298490)
It's definitely absurd, but come on, unconscionable makes it sound like it's more irresponsible than the industry standard - when in fact it is less.

My guess is that they are close to shipping 10.9.2 and decided to hold on for that, not realizing how much attention this bug would get.

Yes, I agree that's a stupid call to make, but most smartphones have far worse unpatched vulnerabilities.

Also, it's worth saying that negative attention probably has little effect on Apple. In this case it's totally justified and well deserved, but the vast majority of it is 'they rejected my flappy bird app', or 'why Apple should fire Tim cook', so this doesn't add much to the pile.

Re:Has goto fail been fixed yet? by discostrings on Tuesday February 25, 2014 @12:55PM (#7299185)
> unconscionable makes it sound like it's more irresponsible than the industry standard - when in fact it is less.

> Yes, I agree that's a stupid call to make, but most smartphones have far worse unpatched vulnerabilities.

That's why Apple's response here is so interesting: it's much more common for smartphones to have serious and known unpatched vulnerabilities than it is for desktop and server operating systems to have them.

It's expected that a smartphone might go unpatched for a while, but the industry standard is that a desktop or server operating system should receive a patch for an extremely severe security vulnerability almost immediately.

Why Apple has chosen to ignore and invert that expectation is unclear. Was there an active, seriously damaging attack against iOS that they thought needed to be stopped right away, despite the cost to OS X users? Did they look at the install base and decide patching iOS first would have a larger impact? Is their development process for OS X not up to the task? Do they just care a lot more about their consumer electronics than their computers?

Then there's the question of why they published the details of the iOS patch before publishing a patch for OS X. Perhaps they rushed to patch iOS when they discovered the vulnerability without realizing that OS X is affected as well? The situation raises a lot of questions--while I'm not sure that what has happened is unconscionable yet (rather than merely extremely incompetent but well-intentioned), it's still an open question, and it's certainly possible.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @11:29AM (#7298533)
Do you have examples of these far worse unpatched vulnerabilities? Not saying they aren't there, but I'd like to know what they are.

Anyway, I stand by "unconscionable" either way. It implies nothing about the industry standard. If Apple's unconscionable actions here are still better than the industry standard (and I'm skeptical about that, hence the "if", but even given that) then that just means the industry standard is unconscionable.

Why is it worth saying that negative attention probably had little effect on Apple? I really could not possibly care less about this. My statements are based on impact to the users including myself, not the price of AAPL or its balance sheet.

Re:Has goto fail been fixed yet? by jamiesonbecker on Tuesday February 25, 2014 @11:49AM (#7298664)
Apparently, there ARE some things $100M in cash can't buy, like security reviewers, security teams, security coders.. hmm, actually, $100M in cash could buy all those things. Guess it's a question of priorities, and security isn't one of them.
Re:Has goto fail been fixed yet? by daeken on Tuesday February 25, 2014 @12:42PM (#7299084)
Even if they spent every last cent of that money on security, they would still have bugs. You never get rid of all of them.

People that talk about security spending miss the inherent imbalance in security: a defender has to find every bug; an attacker only has to find one.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @03:22PM (#7300198)
That doesn't mean you can't make it harder for the attacker.

Nobody is saying that Apple's wealth should mean they have no bugs. However, it does mean that they should have reasonable test coverage of critical security code.

Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @07:53PM (#7301952)
Or even if not test coverage - static analysis.
Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @11:57AM (#7298728)
You mean $100B in cash.

Yes - it's obvious that there are bad practices at work here - in particular no strict static analysis. I agree there's little excuse for not adding machine processes that could have helped.

But adding human processes costs time and agility, and as you point out, money cannot replace these.

Re:Has goto fail been fixed yet? by ballard on Tuesday February 25, 2014 @12:39PM (#7299060)
British million = American billion = 10^9

MM is used to denote an American million = 10^6

Re:Has goto fail been fixed yet? by omh on Tuesday February 25, 2014 @12:44PM (#7299099)
Are you sure? The old 'British Billion' was 10^12, but even that is rarely used any more. I've never seen an 'M' suffix used to mean anything other than 10^6.
Re:Has goto fail been fixed yet? by rdl on Tuesday February 25, 2014 @02:34PM (#7299876)
M is widely used as "thousand" in specialized contexts like "CPM" and "RPM" within the advertising/accounting space.
Re:Has goto fail been fixed yet? by Toucan on Tuesday February 25, 2014 @01:24PM (#7299387)
I don't think that's right. 10^6 is a million in both languages.

10^9 used to be a milliard in British English but is no longer used. It is now referred to as a billion.

10^12 used to be a billion in British English but is now referred to as a trillion.

Edit: http://www.youtube.com/watch?v=C-52AI_ojyQ explains it all quite nicely.

Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @11:55AM (#7298708)
Here's an example: http://m.slashdot.org/story/198449

And these android vulnerabilities are essentially unpatchable by design - which I think is something that really does deserve the word unconscionable.

But I accept that the overall behavior in the industry could be described by that term.

My comment about the negative attention is that Apple makes decisions based on the impact to itself. A few days of delay in fixing the bug might result in some small number of compromises, but as long as it doesn't drag on for months, it's unlikely to be a big deal.

On the other hand, negative attention and social media sentiment could have been a big deal for Apple, had it not been for all the wolf-crying. Now it's just a muted signal.

So from Apple's position this is actually a minor affair. We are horrified in principle but they are basing their actions on their practical reality.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @01:41PM (#7299505)
Sheesh. I retract my "skeptical about that" bit.

I think you're right that this won't be a big deal for Apple. It won't really hurt them much.

However, basing your actions solely on the impact to yourself and ignoring the potentially huge negative impact for your customers is exactly the sort of thing that causes me to call it "unconscionable".

Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @02:34PM (#7299873)
I agree that if it were just about the impact to themselves this would be "unconscionable", but I think that in practice the effect on users will also be very limited unless this vulnerability has already been known and exploited in the wild, or unless they take weeks more to patch OSX. Remember it requires a MITM attack to exploit it - not just a compromised host.

There is definitely a gamble they are taking, and definitely process improvements they should make (e.g. mandatory static analysis of shipping code). I just think think this is more about poor performance than poor morals.

[noting again that if this was discovered because it's in the wild, then it's all about performance]

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @03:19PM (#7300182)
The iOS patch was released on Friday, so it was publicly known then at the latest. It's not particularly hard to exploit, so it seems entirely reasonable to think that people were exploiting it soon after.

How many Macs connected to Starbucks WiFi this weekend and had their e-mail and banking credentials lifted? If my hat were darker, I'd have been out there doing it for fun if nothing else.

The fact that it requires a MITM means you're fairly safe on a home connection, but there are lots of people out there on public WiFi who would be pretty vulnerable to this.

I really don't understand what the "gamble" is. What's the upside to waiting?

Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @07:52PM (#7301948)
I think your example of a Starbucks is a good one, and I'd be curious to know. The attacker has to go and physically sit in or near the Starbucks to do it, which is a strong limitation. They also have to have to have something worthwhile to do with what they steal. So that limits it to criminals with the knowledge to exploit this MITM attack who are willing to sit outside a Starbucks for long enough to harvest a worthwhile set of credentials.

I doubt that many credentials were stolen this way.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @08:45PM (#7302188)
But some were, right?

Again, what's the upside to waiting? It looks like zero to me. Minor convenience for Apple at best. Placing their own convenience over the security of their customers is bad.

Re:Has goto fail been fixed yet? by ZoFreX on Tuesday February 25, 2014 @01:35PM (#7299462)
> Do you have examples of these far worse unpatched vulnerabilities? Not saying they aren't there, but I'd like to know what they are.

Search for what you can do for an Android phone that isn't on the very latest version of the OS, and then look up stats of how many Android phones are not, and will never be, on the latest version of the OS. That's unconscionable. To not backport fixes to devices, and to stop shipping updates to them mere months after their release in some cases.

Re:Has goto fail been fixed yet? by allochthon on Tuesday February 25, 2014 @11:20AM (#7298455)
I'm a little naive when it comes to SSL/TLS. I've been wondering whether the reason for the delay is that because with this compromise Apple's update service is no longer a secure channel through which to distribute a fix. So now they're scratching their heads trying to figure out a way around the issue, possibly coding up something that uses OpenSSL. Is this line of reasoning unfounded?

EDIT: Great points about the checking of the signatures. Let's hope there's not a second bug that can bypass this in some cases.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @11:22AM (#7298479)
My understanding is that the update system checks digital signatures on the downloaded data separately from TLS, rather than simply relying on the integrity of TLS. If that understanding is correct, then there shouldn't be any issue there.
Re:Has goto fail been fixed yet? by gress on Tuesday February 25, 2014 @11:27AM (#7298518)
That would be all the more reason for them to have released the update simultaneously with the iOS update.

Although we still don't know how this was discovered or if it's in the wild.

Re:Has goto fail been fixed yet? by aroch on Tuesday February 25, 2014 @11:26AM (#7298516)
Software updates are signed by Apple's key, only signed and verified files can be installed by software update
Re:Has goto fail been fixed yet? by bct on Tuesday February 25, 2014 @11:26AM (#7298517)
That's an interesting line of thought, but:

1. if they can't use the update service, then they can't deploy anything that they're coding up anyways.

2. I assume they sign their updates, so it shouldn't matter that the channel is compromised.

Re:Has goto fail been fixed yet? by seeken on Tuesday February 25, 2014 @11:17AM (#7298435)
I wonder if they are finding that a lot of their systems inadvertently relied upon this bug, and they are scrambling to test and fix them all rather than release a fix that causes a lot of other things to break
Re:Has goto fail been fixed yet? by beat on Tuesday February 25, 2014 @11:21AM (#7298461)
More likely they have to regression test a whole ton of things, just in case, and do all the recompiling in correct order. I can't see anything actually depending on the bug, but making sure they don't screw up the patch is hard.

But a lot of programmers who have never done anything more difficult than mylamesocialstartup.com in PHP have no idea what it's like to build and test something as complex as an OS. No, recompiling your Linux kernel ain't the same thing.

Re:Has goto fail been fixed yet? by 0x0 on Tuesday February 25, 2014 @11:24AM (#7298496)
If they need more than a day to recompile one library with a one-line change and see if it still boots and runs software update, then they have much bigger issues than this single bug.

It's now been 5 days PLUS however long they sat on this for ios.

Re:Has goto fail been fixed yet? by beat on Tuesday February 25, 2014 @11:59AM (#7298745)
They do have a bigger issue. The bigger issue is that they're releasing an entire operating system, with 15 years or so of accumulated cruft and process. And it's not just a one line change... they need to understand the implications of the bug across their system and create regression tests to catch all those corner cases, then do the "one line change" and properly regression test it everywhere.

Again, this isn't some little web site. This is a piece of software that thousands of programmers have worked on for more than a decade, millions of lines of code with a tremendously complex build, testing, release, and approval system.

And finally, Apple's primary concern isn't that a handful of customers might be exposed to risk for a couple of extra days. It's that they botch the rollout of a major security patch and have to re-release, or customers are victimized by new bugs as part of the patch. That sort of thing can cost them billions in market punishment.

Re:Has goto fail been fixed yet? by 0x0 on Tuesday February 25, 2014 @12:06PM (#7298793)
What? It's a one-line, clearly understood bugfix. (The extra goto wasn't there in 10.8.x and that hasn't caused any problems) Rebuild the library, spend a few days regression testing at least software update so you can re-push another update, and ship it at the same time as the ios and appletv fixes!

You don't spend a week holding the fix hostage so you can fine tune a new os release with "improvements to autofill forms" in safari while script kiddies are running wild with mitmproxy.

Also I would think that a full ios+appletv OS release is much more complicated than an OSX release due to the way the mobile OSes are packaged (probably 10+ unique/model restore images plus delta downloads)

Re:Has goto fail been fixed yet? by jamiesonbecker on Tuesday February 25, 2014 @11:53AM (#7298696)
"No, recompiling your Linux kernel ain't the same thing."

You're right. It's not at all like Red Hat compiling and releasing a new RPM which they would push out in hours, max. It's actually much easier. Apple retains absolute control over their source code.

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @11:25AM (#7298500)
It's not like iOS is some tiny little thing, but they got it out much sooner.

If their build process is so broken that it takes days to take the 10.9.1 tag, apply a one-line patch, and release it, then they're doing it severely wrong. Security problems happen, and you need to be ready to move fast.

Personally, if I was on that team and it came down to taking days to recompile and retest everything, I'd be seriously considering a binary patch as an interim fix. Take the actual built binaries, apply this one patch, and you know nothing else got somehow miscompiled or mislinked in the process.

Re:Has goto fail been fixed yet? by rdl on Tuesday February 25, 2014 @11:26AM (#7298513)
Most likely it's that they're trying to do the patch in 10.9.2, and have a lot of non-bug-fix related 10.9.2 stuff to work out (they've done 7 prerelease builds to developers of 10.9.2 already).

The manager who is holding the SSL fix hostage to 10.9.2's shitty feature-adds should be terminated out of a cannon. It's hilarious that we don't know who that person is -- who is the head of software security at Apple? From what I've heard there are at least two separate teams, and the relevant team is closely tied to Federighi, but no one outside Apple (generally) knows sub-SVP people at Apple.

There should be 10.9.1.1 with just the SSL fix. If that means delaying 10.9.2 another month, so be it.

Re:Has goto fail been fixed yet? by 0x0 on Tuesday February 25, 2014 @11:19AM (#7298450)
Wonder if we'll see a patch for this 0day this side of WWDC...
Re:Has goto fail been fixed yet? by ballard on Tuesday February 25, 2014 @12:36PM (#7299040)
Apple lacks an emergency fix process. Also the fix will be released in 10.9.2, but it's held up in testing amgonst all the new added functionality.
Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @01:42PM (#7299513)
I've seen quick reactions to security problems before. iOS didn't seem to have the same trouble. And if they lack an emergency fix process, that in and of itself is unconscionable.
Re:Has goto fail been fixed yet? by sneak on Tuesday February 25, 2014 @02:04PM (#7299668)
> why it's taking so long to fix it on the Mac side

So, it turns out that they had a point release very far down the release engineering pipeline that included several new features (FaceTime Audio, FT call waiting, Mail bugfixes, et c).

A few days ago when the TLS bug became widely known they had two options: push a point release for _just that_ near-immediately, followed by the existing planned release (after redoing QA because the TLS changes would require the feature-add release be rechecked), or just integrate the TLS fix and restart QA on the planned release. They opted for making everyone wait a few days and just doing one release.

It was a trade-off, and both would have been okay decisions IMHO. There is a non-zero downside to releasing OS updates too frequently with an installed base as large as OSX.

"Unconscionable" it would have been if it had dragged on for a week or more. This was probably a heroic effort by the release engineering team(s) to get it out in in the timeframe that they did (after opting for just one release).

Re:Has goto fail been fixed yet? by mikeash on Tuesday February 25, 2014 @02:28PM (#7299829)
I completely disagree with your assessment of the two choices. Releasing an immediate security update independent of 10.9.2 was the only reasonable course of action. Holding off on a significant patch for days just because it doesn't suit your release schedule for unrelated stuff is ridiculous.
Re:Has goto fail been fixed yet? by rdl on Tuesday February 25, 2014 @02:33PM (#7299870)
The only excuse for 10.9.2-only would have been if rolling a 10.9.1.1 would have actually taken longer to get out the door at all. But if that's the case, they have serious problems.
Re:Has goto fail been fixed yet? by zeckalpha on Tuesday February 25, 2014 @11:32AM (#7298562)
My guess is that since on iOS every "browser" uses mobile Safari as a backend it was critical. On Mac, Chrome and Firefox do not have this issue, so it isn't as much of an issue. I'm not sure what percentage of Mac users use Safari.
Re:Has goto fail been fixed yet? by 0x0 on Tuesday February 25, 2014 @11:36AM (#7298586)
Yes, nobody cares about their email or icloud account passwords being exposed.

Good thing they made sure all our AppleTVs are safe from MITMs, though.

Re:Has goto fail been fixed yet? by jamiesonbecker on Tuesday February 25, 2014 @11:57AM (#7298730)
Are you saying that since no one actually uses Safari it doesn't matter?

Or are you saying that this doesn't impact the underlying SSL library for all of OSX?

Or are you saying that no one (should) use Mac for actual work, so it doesn't matter?

Obviously, you're not saying these things, but what if you did. :)

Re:Has goto fail been fixed yet? by zeckalpha on Tuesday February 25, 2014 @02:17PM (#7299764)
The vulnerability didn't affect Chrome and Firefox, or my brew-upgraded curl. I wasn't too worried.
Re:Has goto fail been fixed yet? by mandalar12 on Tuesday February 25, 2014 @11:23AM (#7298485)
I am also very surprised by the lack of outrage in media, either mainstream or even tech related (including blogs and HN / netsec).
Re:Has goto fail been fixed yet? by rdl on Tuesday February 25, 2014 @02:31PM (#7299860)
The best data point you'll see is ex-Apple people basically not defending the company, and any Apple departures in the next weeks/months.