the disclosure debate is the wrong argument

prompted by an exchange on twitter this morning i began spending mental resources on an old favourite that everyone loves (hates) to go over again and again - the disclosure debate.

the disclosure debate is a debate over what the right answer is to the question  "i found a problem, now what do i do with it?". it's a natural place to come from for many in the security community and the question surely needs an answer. but the disclosure debate tries to raise the answer into a principle and each side of the debate believes that their side is more principled, more virtuous than the other. you have to pick a side and in so doing you also pick a tribe and signal your allegiance.

the disclosure debate has been raging in the security community for a long time and the two major sides right now are full disclosure which is generally favoured by security researchers, and responsible disclosure which is generally favoured by vendors. when you pick a side you're choosing to align with one of these groups.

but is that really it? is it just a choice between researchers and vendors? are they the only ones with real skin in the game? hell no.

so then let's catalog the groups of people involved:

1. vendors of vulnerable products/systems/services. these are the people who, when faced with discovery of a security vulnerability in their product, are expected to act quickly and create a fix for their product to eliminate the risk posed by by that vulnerability
2. security researchers. these are the people who find the vulnerabilities. often a great deal of work goes into finding vulnerabilities and consequently researchers often want and even deserve some kind of reward, whether simply adding to their reputation which they can parlay into better paying job or a more direct financial reward like a bug bounty.
3. attackers. these are the people who would actually exploit the vulnerability to victimize people, whether for money, ideology, or a laundry list of other reasons
4. users of vulnerable products/systems/services, be they businesses, institutions, individuals, etc. these are the people who are actually affected by vulnerabilities and arguably should be the most important consideration

now attackers and users don't figure all that much in the disclosure debate. sure the ostensible reason we have the debate is because we say we want to help the users, and the reason for doing the vulnerability research and disclosing the results at all is supposed to be to thwart attackers, but when arguing over which side of the disclosure debate is best attackers and users are treated as constants, as monolithic homogeneous groups that always present the same issues in every case and that's just not an accurate reflection of reality. 

not all attackers would be able to find the vulnerabilities on their own. those that can are just as much a minority as the security researchers themselves are. not all attackers are interested in any given vulnerability. those that target foreign states may be more interested in PLC vulnerabilities than VLC vulnerabilities. some may not have the connections or resources to exploit a particular vulnerability. some may specialize in attacks against a completely different kind of technology. some are simply not motivated enough to add a new exploit to their arsenal.

meanwhile, not all users pay attention to the channels where disclosures are communicated, in large part because most users can't understand or meaningfully act on the information in a disclosure. filtering out attack traffic, using IOCs, deploying mitigations, etc. are all the domain of a very select group. additionally not all users are aware of all patches when they become available, or are willing or able to apply them.

these are only a fraction of the variations among the two groups that can affect if/how they'll use a vulnerability and if/how they might be affected by the vulnerability. the reason we keep having the disclosure debate is because that debate has never adequately addressed these variables, because they have the potential to make the answer change depending on the vulnerability, which would mean there is no single right answer or right side in the debate. the debate focuses on principles to the exclusion of nuance, only paying lip service to the people who would use the vulnerabilities and the people those vulnerabilities would be used against. 

those who find a problem in some kind of technology want an easy answer to the question of what they should do with it - that's why the disclosure debate exists, to try and arrive at an easy answer so people can just follow the one right principle for handling the situation and be done with it. it follows a trend that has become all too common in the technology sector of trying to abstract people out of the equation because people are messy and complicated. but people are part of the equation whether we like it or not so the argument shouldn't be about which principle is right, but rather whether our actions should be governed by principle or by considerations for people.

in the watchmen, the character rorschach was governed by principles to the exclusion of virtually everything else ("never compromise. not even in the face of armageddon"). defending truth and justice so stridently made him arguably the most heroic, in the comic book sense of the word. but the more realistic portrayal of individuals and circumstances and society as a whole in the watchmen highlighted how ill-fitting such unwavering adherence to principles can be in the real world.

we need to stop talking about the disclosure debate, not because it should be settled already, but because it can never be settled, because it's parameters don't fit the real world. it's the wrong argument. if we really want to help protect one group of people from another group of people then we need to spend a lot more time thinking about those people each and every time the subject comes up. there are no shortcuts when it comes to doing the right thing.

winblows update

so i'm using my computer thursday night when i get a notification that updates have been applied that require a reboot. ok, whatever, i've mostly come to terms with the fact that Windows 10 updates itself without asking me, at least it asks me when is a good time to reboot. that wasn't a good time so i said later.

well, later came in the wee hours of the morning when i was done working and ready to head to bed. i go to shut off the computer and the new options in the shutdown menu remind me that there was an update to take care of, so i choose the update and reboot option.

i chose that option rather than update and shut down because, from past experience, the update process has not actually completed by the time the system shuts down. there's a bunch of stuff it needs to do on the next restart and that takes up time i could be doing something else, so no shutdown just yet, just an update and reboot and i head off to take care of my evening oral hygiene routine thinking that the process would be done by the time i get back.

when i come back the computer is still in the process of rebooting? what the heck? oh no, i've seen this before (or at least i think i have). is the computer stuck in a reboot loop? powering off for a few moments usually breaks out of the loop, but this time i discover it wasn't a reboot loop at all. when i let it power up again i see a screen saying that it's applying updates and that several reboots will be required.

are you kidding me? this is not what i want to deal with in the wee hours of the morning when i still have to go to work the next day. this is not convenient, and frankly "several reboots" for an update is bullshit. i understand the need to perform a reboot during an update; files and other resources that need to be changed may be locked by running processes and rebooting eliminates that impediment, but several reboots? Microsoft has been at this update business for decades now, you'd think their little minions would have figured out how to coordinate their efforts so that each part of the update could make use of the same reboot, but no, apparently that kind of unified effort is beyond them and in fact they seem to be moving in the opposite direction where every bit and piece of their updates (and the operating system itself) is becoming more separate and isolated from the others.

so i did what i hate doing. i left the computer on completely unattended overnight so that hopefully by morning the update would be done. and it was, but that's not the end of the update related problems. you see, Microsoft's updates aren't just for security fixes. those are important, yes, and the fact that people were taking too long to apply them and leaving their systems to become part of massive botnets is part of the reason the user's control over updates was taken away from them. however, Microsoft has re-imagined how versioning of their operating system will work so those updates now also come with feature changes, which (due to the increasingly isolated approach units within Microsoft are taking nowadays) means new binaries with new behaviours.

how the hell is anyone supposed to develop a behavioural baseline for their system with this never ending parade of new binaries and new behaviours? this morning's culprit? BackgroundTransferHost.exe. what does it do? who the hell knows? not only does Microsoft give us less agency now that we can't control if/when updates occur, but there's also less transparency now too because the number of separate/isolated binaries they're introducing to the system has far, far outpaced anyone's efforts to document them.

maybe BackgroundTransferHost.exe isn't even Microsoft's. maybe it's malware. if i were going to make a downloader trojan, that sounds like just the sort of name i'd use - but what do i know, i'm not a malware writer. i suppose they expect me to trust it because it's signed, but that's not how that works. being signed (and passing the signature validation procedure) just means it hasn't been modified after getting signed, not that it's legitimate, not that it's safe, not that it's trustworthy. signing certificates get stolen. there's plenty of signed malware out there.

oh, and the cherry on top is now VMware is non-functional.

what the actual f#$% Microsoft. stop making alternative security approaches so much harder than they have to be. i'm regretting moving on from Window XP. at least there i could perform application and behavioural whitelisting with relative ease.

thoughts on attack automation

axiom: there is no perfect security

because there is no perfect security we can say with certainty that systems will never be perfectly secure. if we close one vulnerability there will always be another to take it's place. we can spend an unending amount of money/time/effort on closing vulnerabilities and still remain vulnerable. in the process of doing this we would go bankrupt because no one, not even the largest companies in the world, have unlimited resources to spend on security.

therefore, eliminating all the vulnerabilities is not a viable strategy. instead a promising alternative is to approach the problem of security from an economic standpoint. while we can't eliminate all the vulnerabilities, we can eliminate some, and if we eliminate the ones that are easiest to exploit then an attacker's job becomes harder and more expensive to carry out. if we make the attacker's job hard enough then the value/benefits they derive from succeeding in their attack (success will always be possible) would no longer cover the cost of launching that attack.

attack automation doesn't make an attacker's job harder. quite the opposite in fact, it makes it easier. attack automation is carried out by attack tools. as a general rule, attack tools reduce the complexity of performing an attack. tools automate the fiddly bits to save an attacker time and effort but in so doing also save the attacker from needing to know how to do the fiddly bits themselves. this means that a larger population of attackers will become capable of carrying out a particular attack because the technical complexity of performing the attack is reduced. it also means there is a larger pool of targets to victimize because the lower cost of performing the attack makes attacking lower value targets economically viable.

additional automation to save the attacker time and effort when selecting targets and launching attacks is also possible, as the recent release of autosploit has highlighted. this lowers the cost of scaling up the attack so that a single attacker can attack a larger group of victims at a lower cost.

the argument can be made that attackers are entirely capable of making these automated attack tools themselves so it doesn't matter if security researchers do it as well. however, when researchers make the automated attack tools, not only do attackers enjoy cost savings with respect to launching attacks, they also enjoy cost savings with respect to developing the tools to launch attacks.

all of these cost savings for the attacker work against defenders. when the cost of performing an attack is reduced it means that attacks that didn't need to be defended against before (because they were too expensive to launch relative to their payoff) must now be defended against, and that increases the costs for defenders because it requires more to be done.

these cost savings are also permanent. attacks don't get harder, they only ever get easier. offensive security researchers are permanently changing the economics of mounting various attacks in the attackers' favour in an effort to incentivize defenders to do what the researchers think should be done to chase after the zero-vuln goal (a goal which we already know to be unattainable) without regard to the economic realities those defenders face.

enforcing their will, their vision of what security should be on others is misguided and damaging. there is no one-size-fits-all approach to security, and where the fit is bad there will be undue burdens with respect to cost and/or unfortunate breaches that might legitimately have been avoided if attackers had not been given a helping hand.

if attackers can build the tools that make their lives easier by themselves, then let them do it. make them pay the cost of doing it. stop subsidizing attackers in the name of security research. years ago the security research community embraced the idea that there should be no more free bugs - why then are the cybercriminals still getting bugs, and exploits, and frameworks, and more for free after all this time?

thoughts on "In Search of Evidence-Based IT-Security"

Christopher Soghoian brought to my attention a video of a talk by Hanno Böck at the 33rd Chaos Communication Congress. in it Hanno puts forward the claim that IT security is largely science-free, so let's follow a staple of the scientific process - peer review.

Hanno introduces himself as a journalist and hacker and says that he prefers to avoid the term "security researcher" and that he hopes the audience will see why.  for those who are relatively well versed in the field of anti-malware it should definitely become obvious why he prefers to avoid that term and i'll return to this near the end.

Hanno is a skeptic, and far from the only one, his talk ultimately expresses the same sentiments that are now common-place in the perennially misinformed information security community. the difference is that Hanno has found a novel way of expressing them, couched in scientific jargon and easily mistaken for insight. he spends altogether too long and dives too deeply into the medical analogy upon which computer viruses and by extension anti-virus software is named. the analogy has long been recognized as deeply imperfect and limited. that's why, in reality, there are relatively few references to this analogy in the anti-malware field other than "computer virus", "anti-virus", and "infection" (all three of which date back virtually to the beginning of the field). his call towards the end of his talk for blinded or even double blinded studies, aside from being prohibitively expensive to perform, seem to cling to this medical paradigm in spite of the fact that the subject of such experimentation (ie. the computer, since we're interested in whether AV can prevent computers from becoming compromised) cannot be psychologically influenced by knowledge of which (if any) anti-virus is being used.

when he FINALLY leaves the topic of medical science to return to security products (about 14 minutes into his half hour talk) he harps on the absence of one very particular kind of experiment being performed on security products - what he calls a randomized controlled trial. it turns out this is a hold-over from his preoccupation with medical science. when Hanno says that IT security is largely science-free it is the absence of this particular kind of scientific experiment that he is referring to, but that doesn't actually make it science-free because science has a variety of different ways to study and experiment on things that aren't people.

there is in fact good scientific evidence for the efficacy of anti-virus software and it's provided by none other than Microsoft:

now it's true that this is data is from an observational study and that it only shows correlation rather than causation, but that's not the end of the world. observational studies are still science. showing correlation may not be definitive evidence but it's still strong evidence, especially considering the scope of the study (hundreds of millions of computers around the world out of a total estimated population of 1.25 billion windows PCs). in this particular case A may not be causing B but B definitely can't cause A and if anyone can think of a confounding variable that might be present on hundreds of millions of systems then maybe let Microsoft know so that they can try to account for it in the future.

another source of scientific evidence (oft derided in information security circles because the results don't match experts' anecdata) are the independent testing labs like av-test.org or av-comparatives.org. they eliminate the influence of confounding variables and so are capable of showing causation rather than just correlation. unfortunately Hanno believes their methodology is "extremely flawed". let's look at his complaints:

• "If a software detects a malware it does not mean it would've caused harm if undetected."
• this is trivially false. anyone who actually reads the testing methodology at av-comparatives (for example) can find right at the beginning a statement about first testing the malware without the AV present and eliminating any that don't work in that scenario. therefore every sample that is detected by AV in their tests would have caused harm if it had gone undetected.
• "Alternatives to Antivirus software are not considered." (the talk gives "regular updates" and "application whitelisting" as examples)
• the example of "regular updates" is frankly a little bit bizarre given Hanno's earlier references to confounders. not controlling for this scenario would actually introduce a confounding variable and make it more difficult to show a causal relationship between the use of a particular AV and the prevention of malware incidents.
• the example of "application whitelisting" underscores a serious problem in Hanno's understanding of what he's critiquing. application whitelisting isn't an alternative to AV, it's a part of AV. many products include this as a feature. Symantec's product, for example, has what they call a reputation engine which alerts when it encounters anything that doesn't have a known good reputation (which means new/unknown malware, traditionally the bane of known-malware scanning, will get alerted on because it hasn't been seen before and thus no reputation, good or bad).
• "Antivirus software as a security risk is not considered."
• when malware exploiting vulnerabilities in anti-virus software is found in the wild then perhaps the test methodologies should be updated to include this possibility. until then, changing the methodology to account for malware that doesn't seem to exist outside a lab has no real benefit.
• "None of these tests are with real users."
• again, this would introduce a confounding variable. maybe the lack malware incidents is because of something the user did rather than because of the AV. alternatively maybe the failure to stop malware incidents is because of something the user did rather than because of a failure of the AV. if you want to establish causation you have to control your variables (something our scientifically-minded speaker Hanno should know all too well). does the anti-virus prevent malware incidents? the tests say yes. can a user preempt or compromise that prevention? also yes. is there any prevention a user can't preempt or compromise? sadly (or perhaps thankfully) no. if you want a study that includes users and thus eliminates the ability to establish a causal link between AV use and prevention of malware incidents, see the study by Microsoft, but even with the inclusion of the users it still suggests AV prevents malware incidents.

when Hanno addressed the paucity of scientific papers dealing with security i found myself confused. using Google Scholar to find the most cited scientific papers? surely he doesn't think the realm of security is so narrowly focused that he'll find what he's looking for that way. security is in fact incredibly broad, covering many different quasi-related domains, and looking at a handful of the most popular scientific papers across all of security is in no way representative of the corpus of available works related to any one particular field (like security software). perhaps i'm biased, having previously (in the very distant past) maintained a reference library of papers related specifically to anti-virus, but it doesn't seem like Hanno showed much evidence that he knew how to find evidence-based security. is it really that hard to add the term "malware" to his search query? could he not find a few and then use them as a seed in an algorithm that crawls backwards and forwards through scientific papers by citation? did he even bother to look at Virus Bulletin? does he even know what that is?

security isn't the only thing that is incredibly broad - so too is the practice and discipline of science itself. there are many different fields and each one does things in their own particular way. we do not perform randomized controlled trials on the cosmos. as a general rule we do not intervene in volcano formation. the work being done at the large hadron collider does not follow exactly the same methodologies that are used in medical science. are we to judge cosmology, volcanology, or particle physics poorly because of this? no of course not. a question you might well ask is what kind of science should logically be used when it comes to studying computer security and, while i suspect multiple scientific disciplines could be useful, the one that springs immediately to mind is computer science. does computer science look anything like medical science? as someone with a degree in computer science i can tell you the answer is emphatically no. we do many things in computer science but randomized controlled trials are not among them (because computers are not people). while Hanno may style himself as "scientifically minded" he doesn't seem to demonstrate an appreciation for the breadth of valid scientific research methodologies and one is left to wonder if he's familiar with any kind of science outside of medicine.

when it comes right down to it, it's this apparent lack of familiarity with the subject matter he's talking about that i found most troubling about Hanno's talk.what is anti-virus software really? what is av testing methodology really? what does science really look like? where do you look for scientific research into malware and anti-malware? these all seem to be questions Hanno struggles with, which brings us back to the subject of why he likes to avoid the term "security researcher". if i had to venture a guess i'd say it's because he doesn't do research, even the basic research necessary to understand the subject matter. as such i would say avoiding the term "security researcher" is probably appropriate (for now).

i'm not sure what one can say in a talk about a subject one hasn't done one's homework on, but hopefully that can improve in the future. Hanno referenced Tavis Ormandy during his talk (as people who criticize AV like to do). Tavis' work on AV also suffered from a lack of understanding in the beginning, but he improved over time and, while he still has room for more improvement, now has arguably done some good work in finding vulnerabilities in AV and holding vendor's accountable for the quality of their software. i'm certain Hanno can also improve. i know there are real criticisms to be made of AV software and the industry behind it, but they have to be informed, they have to come from a place of real knowledge and understanding. i look forward to Hanno reaching that place.

highlights from #sector2016

i haven't posted about the sector conference in a number of years, in spite of attending. let's break that trend. here's some highlights from this year's conference.

• edward snowden - he was surprisingly well prepared to talk about canadian policy and events
• marketing gimmicks - hockey pucks are one thing, but give me a key and tell me that there's a lock that it might fit and you're darn right i'm going to go find out if it fits. i gather there was a prize i could have won if it did fit but i didn't pay much attention to that. a friend at work said he'd have bought their product on the strength of that gag alone.
• ransomware - ransomware seemed to be the theme this year. i lost track of how many talks were about or mentioned ransomware. 2016 really does seem like the year of ransomware. i caught the tail end of talk from someone at sophos where they described a feature for rolling back encrypted files using a proprietary backup mechanism. if other vendors aren't doing something like this you're leaving money on the table. (that's right, i know how you vendors think)
• mikko hypponen - great perspective on what protecting computers has evolved into: protecting society because it now runs on computers
• the security problems of an 11 year old and how to solve them - this talk was given by an actual 11 year old who could probably put some professionals to shame. this is the talk i most look forward to sharing with people at work when the videos become available.
• mikko's "virus" floppy disk that he left behind - this isn't a highlight because someone from the AV industry was careless with infectious materials but rather because when it was found people wanted to find a computer they could stick it into and see what was there. you'd think the difficulty in finding the hardware necessary to read a 5 1/4" floppy would make such a disk a relatively safe prop to use, even if there was a virus on it. leave it to infosec pros to try and find ways around such barriers. now you know where shadow IT comes from, folks. by the way, don't tell mikko.

there were other good talks and keynotes, of course, but i'm not going to detail every talk i attended and every person i met. these are the things that really stood out to me and if you want to know more you should have gone yourself.

the anti-virus harm balance

anti-virus software, like all software, has defects. sometimes those defects are functional and manifest in a failure to do something the software was supposed to do. some other times the defects manifest in the software doing something it was never supposed to do, which can have security implications so we classify them as software vulnerabilities. over the years the software vulnerabilities in anti-virus software has been gaining an increasing amount of attention by the security community and industry - so much so that these days there are people in those groups expressing the opinion that, due to the presence of those vulnerabilities, anti-virus software does more harm than good.

the reasoning behind that opinion goes something like this: if anti-virus software has vulnerabilities then it can be attacked, so having anti-virus software installed increases the attack surface of the system and makes it more vulnerable. worse still, anti-virus software is everywhere, in part because of well funded marketing campaigns but also because in some situations it's mandated by law. add to that the old but still very popular opinion that anti-virus software isn't effective anymore and it starts looking like a perfect storm of badness waiting to rain on everyone's parade.

there's a certain delicious irony in the idea that software intended to close avenues of attack actually opens them instead, but as appealing as that irony is, is it really true? certainly each vulnerability does open an avenue of attack, but is it really doing that instead of closing them or is it as well as closing them?

if an anti-virus program stops a particular piece of malware, it's hard to argue that it hasn't closed the avenue of attack that piece of malware represented. it's also hard to argue that anti-virus software doesn't stop any malware - i don't think anyone in the anti-AV camp would try to argue that because it's so demonstrably false (anyone with a malware collection can demonstrate anti-virus software stopping at least one piece of malware). indeed, the people who criticize anti-virus software usually complain not about set of malware stopped by AV being too small but rather that the set of malware stopped by AV doesn't include the malware that matters most (the new stuff).

so, since anti-virus does in fact close avenues of attack, that irony about opening avenues of attack instead of closing them isn't strictly true. but what about the idea that anti-virus software does more harm than good? well, for that to be true anti-virus software would have to open more avenues of attack than it closes. i don't know how many vulnerabilities any given anti-virus product has so i can't give an exact figure of how many avenues of attack are opened. i doubt anyone else can do so either (though i imagine there are some who could give statistical estimates based on the size of the code base). the other side of the coin, however, is one we have much better figures for. the number pieces of malware that better known anti-virus programs stop (and therefore the number of avenues of attack closed) is in the millions if not tens of millions and that number increases by thousands each day. can the number of vulnerabilities in anti-virus software really compare with that?

it's said that windows has 50 million lines of code. if an anti-virus product were comparable (i suspect in reality it would have fewer lines of code) and if that anti-virus product only stops 5 million pieces of malware (i suspect the real number would be higher) then in order for that anti-virus product to do more harm than good it would need to have at least one vulnerability for every 10 lines of code. that would be ridiculously bad software considering such metrics are usually estimated per 1000 lines of code.

now one might argue (in fact i'm sure many will) that those millions of pieces of malware that anti-virus software stops don't really represent actual avenues of attack because for the most part they aren't actually being used anymore. they've been abandoned. counting them as closed avenues of attack isn't realistic. the counter-argument to that, however, is to examine why they were abandoned in the first place. the reason is obvious, they were abandoned because anti-virus software was updated to stop them. the only reason why malware writers continue making new malware instead of resting on their laurels and using existing malware in perpetuity is because once anti-virus software can detect that existing malware it generally stops being a viable avenue of attack. so rather than the abandonment of that malware counting against anti-virus software's record of closing avenues of attack it's actually closer to being AV's figurative body count.

there is still malware out there that anti-virus software hasn't yet stopped, and as that set is continually replenished it's unlikely that anti-virus software will stop all the malware. it has stopped an awful lot so far, however, so the next time someone says anti-virus software does more harm than good (due to it's vulnerabilities) ask them for their figures on the number of vulnerabilities in anti-virus products and see how it compares with the number of things anti-virus software stops. i have a feeling you'll find those people are full of it.

there's a quality problem in the anti-malware industry

if you follow infosec news sources at all, by now you've probably heard about the claim made by an anonymous pair of ex-kaspersky employees that kaspersky labs weaponized false positives.

more specifically, the claim is that engineers at kaspersky labs were directed to reverse engineer competing products and use that knowledge to alter legitimate system files by inserting malicious looking code into them so that they would both seem like files that should be detected and be similar enough to the original file that the competing product will also act on the legitimate file and in so doing cause problems for users of those competing products.

i've heard this described as fake malware, but for the life of me i can't see why it should be called fake. the altered files may not do anything malicious when executed, but they're clearly designed to exploit those competing products. furthermore, there is clearly a damaging payload. this isn't fake malware, it's real malware. it may launch it's malicious payload in an unorthodox and admittedly indirect manner, but this is essentially an exploit.

some consider the detection of these altered files to be false positives because the files don't actually do anything themselves, but since they have malicious intent and indirectly harmful consequences, i think the only real false positives in play here are the original system files that are being mistaken for these modified files.

by all accounts, this type of attack on anti-malware products actually happened. what's new here is the claim that kaspersky labs was responsible at the direction of eugene kaspersky himself. there's a lot of room for doubt. the only data we have to go by so far, besides the historical fact of the attack's existence, is the word of anonymous sources (who potentially have an ax to grind) and some emails that, quite frankly, are easily forged. circumstantially there's also an experiment kaspersky participated in around the same time frame that has similar earmarks to what is being claimed except for the part about tricking competing products into detecting legitimate files as malware.

i don't expect we'll ever know for sure if kaspersky was behind the attacks. doubts have been expressed by members of the industry, but frankly i've seen too many things whitewashed or completely ignored (like partnerships with government malware writers) to take their publicly expressed statements at face value. there are certainly vendors i'd have a harder time believing capable of this but there just doesn't seem to be sufficient evidence that the claims are true. the problem is that i can't imagine any kind of evidence the anonymous sources are likely to have that isn't easy to repudiate. had they taken a stand at the time (like someone with actual scruples would have done) they would have been able to put their names behind their claims - they may have lost their jobs but they surely would have been able to find employment with a different vendor because hiring a whistle-blower would have been good PR.

however, as it stands now, the anonymous sources have to remain anonymous. if they're telling the truth then they are complicit in kaspersky's wrong-doing, and if they're lying they are throwing the entire industry under the bus for no good reason (because this claim fans the fires of that old conspiracy theory about AV vendors being the ones who write the viruses). Either way, to have this claim linked to their real identities now would make them radioactive in the industry. no one would touch them, and for good reason.

long ago it used to be that the industry only employed the highest calibre of researchers. people who were beyond reproach. naturally, in order to grow, the industry has had to hire ever increasing numbers of people and old safeguards against undesirable individuals joining the ranks don't scale very well. increasingly people who aren't beyond reproach are being found amongst the industry's ranks and there appears to be no scenario where these two anonymous sources don't fall into that category. the inclusivity that the general security community embraces (and that the anti-malware industry is increasingly mimicking) has the consequence that blackhats are included. the anti-malware industry is going to have to either figure out if they're ok with that or figure out a way to avoid what the general security community could not.

no malware defeats 90% of defenses

yesterday, 'security expert' robert graham penned a blog post claiming that all malware defeats 90% of defenses - a claim made in answer to the FBI's claim that the attack on sony would have been just as successful against 90% of other companies. as you might well imagine, however, robert graham was in error.

the error isn't a straight-forward one, but it is one that most of the security industry makes. it's an error in framing.

the security industry likes to frame the problem as automaton vs. automaton because that facilitates the comforting lie they tell their customers. businesses see security (not incorrectly) as something that costs them time and money and so they search for ways to cut those costs. the security industry, flush with skillful sales people, tells businesses what they want to hear: that they can cut costs and automate much of security, leaving only a handful of personnel left to operate a little like janitorial staff - cleaning up messes and keeping the automaton running smoothly. likewise, the security industry tells consumers what they want to hear as well: that they just need to install a product and that product will take care of security for them automatically.

in security, however, your adversary isn't a thing, it's a person. malware doesn't defeat defenses anymore than a pick and tension wrench defeats the tumblers in a lock. malware is an object, not a subject. it may have some small measure of autonomy (some more so than others), but it doesn't defeat anything - it's not the agent in that kind of scenario, it's simply a proxy for an intelligent adversary.

intelligent adversaries are notoriously good at outsmarting automatons. robert graham provided a wonderful example of that in his own post when he described creating brand new malware that went undetected by the anti-malware software being run by his targets. what he failed to do was take appropriate credit. it wasn't the malware the defeated those defenses, it was a person or persons with APT level skill (even if it didn't require quite that much skill to pull it off - he described it as easy, but easy is a relative term). the targets were compromised, not because they were using substandard defensive technology per se, but because they were relying on automatons to protect them against people.

in a battle of wits between an automaton and an intelligent adversary, the intelligent adversary has the advantage by definition.

so long as the security industry continues to tell their customers what they want to hear instead of what they need to know, those customers are going to continue relying on a stupid box to fend off smart people. that is a recipe for failure no matter what technology is involved.

stealing master passwords is just not that big a deal

by now a lot of people have heard the news that a new version of the citadel trojan steals the master password for password management software. a LOT of electrons have gone into reporting this new development over and over and over again, but it really doesn't seem like it's a big enough deal to warrant all this attention.

while it may be novel to use keylogging to steal specific passwords for password management software, password stealers and keyloggers are anything but new, and the biggest difference between having this new version of citadel on your system or a traditional keylogger is basically that citadel will be able to collect (and therefore compromise) all your passwords faster than a normal keylogger (all at once vs. piecemeal).

that's really all there is to it. from the perspective of a potential victim, citadel isn't doing anything really new, it's just doing it more efficiently. password stealing malware has been out there for a long time and password managers were never meant to combat that particular threat to password security. password managers are meant to facilitate the use of strong, unique passwords which in turn serves to mitigate the risk of compromises to remote systems - compromising the local system is an entirely different problem.

at the end of the day, you can't operate securely with a compromised computer. even if you were to use 2 factor authentication (which could conceivably render password stealing moot) everything else you enter or access would be exposed to potential theft or manipulation if you're using a compromised computer.

i realize it may seem awkward that a class of software security pros have been promoting for years in order to improve security is now being targeted by malware, but it's only awkward because such a needlessly big deal is being made out of it. password management software still mitigates the same risks associated with remote compromises that it always did,  and you're as hosed as you ever were in the event of a local compromise. nothing has actually changed for the people trying to keep their things secure so stop acting like this development changes anything - it doesn't.

the PayPal pot calling the Apple Pay kettle black

so if you haven't heard yet, PayPal took out a full page ad in the New York Times trying to drag Apple Pay's name through the mud based on Apple's unfortunate celebrity nude selfie leak. This despite the fact that PayPal happily hands out your email address to anyone you have a transaction with. In essence, PayPal has been leaking email addresses for years and not doing anything about it, so they shouldn't get to criticize others for leaking personal information.

what's the big deal about email addresses? while it's true that we often have to give every site we do a transaction on an email address, we don't have to give them all the same address. in fact, giving each site a different email address happens to be a pretty good way to avoid spam, but more importantly it's a good way to avoid phishing emails, and that's important where PayPal is concerned because PayPal one of the most phished brands in existence.

unfortunately, because PayPal wants all parties in a transaction to be able to communicate with each other, they do the laziest, most brain-dead thing one can imagine to accomplish this: they hand out your PayPal email address to others, which is pretty much the worst email address to do that with. i have actually had to change the disposable email address i use with PayPal because they are apparently incapable of keeping that address out of the hands of spammers, phishers, and other email-based miscreants. furthermore, i also use their service less because i don't want to have to clean up after their mess.

at some point i may have to start creating disposable PayPal accounts and use prepaid debt cards with them. certainly if i were trying to hide from massive spy agencies then that would be the way to go, but if i'm only concerned with mitigating email-borne threats i really shouldn't have to go to that much trouble. there are other, more intelligent things that PayPal could, even should be doing.

• they could share the email address of your choosing, rather than the one you registered with their service unconditionally. that way you could provide the same address you probably already provided that other party when you created an account on their site. it shouldn't be too difficult for them to verify that address before sharing it with the other party since they already verify the one you register with.
• they could offer their own private messaging service so that communication could be done through their servers (which would no doubt aid in conflict resolution).
• they could provide a disposable email forwarding service such that the party you're interacting with gets a unique {something}@paypalmail.com address that forwards the mail on to the email address you registered on PayPal with, and once the transaction is completed to everyone's satisfaction the address is deactivated.

they don't do anything like that, however. here's what you can do right now with the facilities PayPal makes available. it's a more painful and less intuitive process than anything proposed above, but it does work.

1. before you choose to pay for something with PayPal, log into PayPal and add an email address (the one you want shared with the party you're doing a transaction with) to your profile. PayPal limits you to 8 addresses.
2. confirm the address by opening the confirmation link that was sent to that address
3. make that address the primary email address for your account
4. confirm the change in primary email address (if you have a card associated with your PayPal account, PayPal may ask you to enter the full card number)
5. at this point you can use PayPal to pay for something and the email address that will be shared with the other party is the one you just added to your PayPal account
6. once you've paid with PayPal you will probably want to log back into PayPal, change the primary email address back to what it originally was (and confirm the change once again) and then remove the address you added for the purposes of your purchase. the reason you'll likely want to do this is because PayPal sends emails to every address it has on record for you, and those duplicate emails will get old fast.

most people aren't even going to be aware that they can do this to keep their real PayPal email address a secret from 3rd parties. as a result all manner of email-borne threats can and eventually will wind up in what would otherwise have been a trusted email inbox. make no mistake, this isn't PayPal providing a way to keep that email address private, this is a way of manipulating PayPal's features to achieve that effect. there are too many unnecessary steps involved for this to be the intended use scenario.

as such, PayPal is leaking a valuable email address by default every time you pay for something. yes Apple's selfie SNAFU was embarrassing to people, and yes if Apple doesn't do something about that now that they're becoming a payment platform it could be not just embarrassing but financially costly for victims, but PayPal is already assisting in similarly costly outcomes right now (not to mention potential malware outcomes) so they really have no right to be criticizing Apple. Apple, at least, is taking steps to correct their problems - what is PayPal doing?

on the strength of authentication factors

i ran across a story on friday about barclays bank rolling out biometric authentication for online banking and wound up starting a debate on twitter that i didn't have time for and couldn't easily fit into a series of tweets even if i did have time. essentially what it came down to was that i don't believe all authentication factors are equally strong and the statement that the barclays system was a "password replacement" raised a red flag for me.

the reason it raised a red flag for me is because single factor biometric authentication is something i've come across before, and not just in an article on the web or even as a user but as a builder. my first job out of university was with a biometric security company, and one of the biggest projects i had while working there was developing an authentication enhancement for windows logon. one of the requests made (and the one i fought the hardest against) was to allow logon with just the biometric. 

here's the problem with this idea - since windows didn't have biometric capabilities built in, the only way to add single factor biometric authentication in was to store more traditional authentication data that windows could accept (such as a password) and then pass that along to windows when the subject's biometric sample matched the registered biometric template. i should note that the article about barclays makes it clear they'll be doing the same thing since they say that barclays won't be storing customer biometric data on their servers. there will have to be a local biometric client that stores more traditional authentication data and passes it on when a biometric match is achieved. storing credentials is not exactly the safest thing in the world. it's not like you can just store a hash of the authentication data in this scenario because you have to be able to present the original, unmodified credentials to the authentication system.

i balked at the idea of making windows less instead of more secure, but i acquiesced when the decision was made to keep the more secure 2 factor mode of operation (without traditional credential storage) in there as well, along with informing the users that biometric-only logon was less secure. 

it's not just less secure because of the credential storage, though, and this is where the twitter debate on friday ventured into. in the course of that job i had the opportunity to examine multiple biometric systems, such as face, voice, iris, etc. and i came away with 2 realizations: 1) the only biometrics that users will ever accept are non-invasive ones (no one wants sensors stuck into them), and 2) that lack of invasiveness makes it relatively easy to steal biometric samples from users, often without them even knowing. fingerprints can be lifted from anything you touch. recordings of your voice can be made without your knowledge. photographs of faces are ubiquitous and a high enough resolution image will capture your iris pattern. 

other authentication factors like passwords and tokens generally rely on restricting access to the authentication data, often through secrecy. when that secrecy is lost, such as when someone takes a photograph of a door key (which is a kind of token) it becomes relatively easy to reproduce the authenticator and gain access to what was being protected. biometrics, especially non-invasive ones, forgo this secrecy under the mistaken belief that reproducing the authenticator is difficult for biometrics. the reality, though, is that you don't have to reproduce a biometric sample, you only have to create an approximation that is good enough to fool the biometric sensor, which often isn't particularly difficult. optical sensors can be fooled with images, audio sensors can be fooled with recordings, the mythbusters once fooled a capacitance sensor by licking a photocopy of a fingerprint.

now hold on, i hear you say, isn't it also really easy to steal passwords? and isn't reproducing that authenticator the easiest of all? it's certainly true that in practice all kinds of things can affect how easy it is for an attacker to become illegitimately authenticated. for that reason i try to look at the upper bound of the strength of the various authentication factors. how strong is a system under ideal conditions, that is where everything goes right and legitimate parties don't make any mistakes.

for passwords, that ideal situation means that the user doesn't accidentally click on anything that would steal his/her password, doesn't get fooled by phishing sites, etc. in short, the attacker can't get the password from the user. it also means the attacker can't get passwords in transit (because that's been properly secured) or a password database from service provider because no vulnerability is found in their system and their employees are likewise careful to avoid making mistakes. under this ideal situation the attacker's only way to succeed in gaining illegitimate entry is to perform an online brute force attack (no, not a dictionary attack, because the user didn't make the mistake of using something from a dictionary) and they'd have to go slow because the ideal provider would have rate-limited failed logon attempts. now you might say this is unrealistic, people make mistakes, and that's true in practice in the aggregate, but it is possible for an individual to do everything right, and it is also possible for attackers to not be able to find any way to attack the provider in order to get the password database. this isn't how strong password protection always is, but rather the ideal we hope to achieve by making our systems secure and avoiding making mistakes, and sometimes in limited cases this is achieved.

for tokens, let's consider the ideal situation to be comparable to that for passwords but on top of that let's consider the strongest token possible (ie. not a door key). let's consider a token that produces one-time-passwords (without any vulnerabilities that would make those passwords easy to predict) so that even brute force attacks become much harder. on the surface this seems even stronger than passwords, but there's a chink in the armour and apple's recent icloud problems are a good example. tokens can be lost or stolen so there needs to be a way to recover from that problem. while our "ideal situation" precludes our user from losing their token, it does not preclude our service provider from providing users with a way to cope with the loss of their tokens. the strongest way to do this is to provide the user with pre-generated one-time-passwords ahead of time. this can work for an individual user who is careful and doesn't make any mistakes but as we've previously seen our "ideal situation" does not extend to the point of saying all users make no mistakes, so the pre-generated one-time-pads are going to fail for reasons such as never being printed out and put in a safe place, or not being able to get to that safe place because the user is traveling, etc. what's a service provider to do then? so far, their best option might be to use traditional passwords as a fall back, and if they do then the token system becomes only as strong as passwords, because although our ideal user didn't lose their token, the provider can't really know that the user didn't lose it (or worse that it was stolen) and has to accept attempts to use the password fall back. while there is room for tokens to be stronger than passwords, the price is that only ideal users will be able to recover in the event of a lost token, and that price may be more than service providers are willing to accept.

for biometrics, we once again consider an ideal user who does nothing wrong, and an ideal service provider who likewise makes no mistakes. in spite of doing nothing wrong the user's voice can still be recorded, their face can still be photographed (in most cultures since facial covering is relatively rare), etc. simply interacting with the world cannot qualify as doing something wrong or making a mistake. acquiring the information necessary to construct a counterfeit authenticator is easy compared to passwords and tokens because no effort is taken to conceal that information and the cultural adjustments needed to change that are beyond what i think would be reasonable to expect. the difficulty in attacking a biometric authentication system boils down to the difficulty in fooling a sensor (or sometimes 2 sensors as people have tried to strengthen fingerprint biometrics with so-called "liveliness tests"), and that difficulty has been consistently overestimated in the past.

this is why i consider biometrics weaker than passwords - because even when everyone does everything right it's still fairly easy to fool the system. as such, when someone (especially a bank) provides people with an authentication system that replaces passwords with biometrics, i think that should raise an alarm. even at that prior job of mine it was conceded that that mode of operation was more about convenience than it was about security. convenience is a double-edged sword, it can make things easier for legitimate users and attackers alike if you aren't careful. using biometrics in a 2 factor authentication system may provide more security than any single factor authentication system can, but biometrics on it's own? there's a reason some people have started saying that your biometric is your username, not your password. don't replace passwords with it (at least not without having someone present to guard against funny business - which isn't an option for online banking).

i wouldn't bet on it

last year cryptography professor matthew green made a bet with mikko hypponen that by the 15th of this month there would be a snowden doc released that showed that US AV companies collaborated with the NSA. he has since accepted that he lost the bet to mikko, but should he have?

i mentioned to matthew the case of mcafee being in bed with government malware writing firm hbgary and mikko chimed in that hbgary wasn't an AV company and being partners with them wasn't enough to win the bet. aside from the fact that this is the first time after all these years that i've seen a member of the AV industry publicly comment on the relationship between mcafee and hbgary (i guess managing matthew's perception of AV is more important than managing mine), something about mikko's response rang hollow.

one way to interpret the situation with hbgary is to view them as government contractors whom mcafee endorsed, advertised, and helped get their code onto the systems of mcafee's customers (hbgary makes a technology that integrates with mcafee's endpoint security product). that certainly would have given hbgary access to systems and organizations they might have had difficulty getting otherwise. i have no idea if that access was ever used in an offensive way, though, so this line of thought is a little iffy.

another way to interpret the situation is to directly contradict mikko and admit that hbgary is a member of the AV industry. after all, they make and sell technology that integrates into an endpoint security product. they may only be on the fringe of the industry, but what more do you have to do to be a member of the industry than make and sell technology for fighting malware? the fact that they also made malware for the government makes them essentially a US AV company that collaborated with the government in one of the worst ways possible.

i feel like this should be enough to have won matthew green the bet, at least in spirit, but the letter of the bet was apparently that a snowden doc would reveal it and the revelation about mcafee and hbgary actually predates snowden's leaks by a number of years. 

so, the question becomes are there any companies that happen to be members of the AV industry and also happen to have been fingered by a snowden leak? it turns out there was (at least) one. they were probably forgotten because they're not just an AV vendor, but AV vendor does happen to be one of the many hats that microsoft wears (plenty of security experts were even advising people to drop their paid-for AV in favour of microsoft's offering at one point in time), and microsoft was most certainly fingered by snowden docs. the instances where microsoft helped the government may not have involved their anti-malware department, but the fact remains that a company that is a member of the AV industry was revealed by snowden documents to have collaborated with the government.

i imagine mikko could find a way to argue this doesn't count either - i admit it's not air-tight - but given how close it meets both the spirit and (as i understand it) the letter of the bet, i think mikko should match the sum he had matthew pay to the EFF and pay it to an organization of matthew's choosing. i won't bet on that happening, though.

confessions of a twitter worm victim

as some of you may know, this past wednesday someone released a self-retweeting worm on twitter that exploited an XSS vulnerability in the popular twitter client tweetdeck. i happen to be a tweetdeck user and i got hit by the worm, not once but twice. since i believe in owning up to my mistakes in order to serve as an example to others, i figured it was important for me to write this post.

this isn't the first time i've had to do this. four years ago it was discovered that there had been a RAT bundled with the software for a USB battery charger sold by the energizer battery company (it had gone undetected by the community for years) and i wrote about my experience then as well.

this was the first time getting hit with something that could spread to others, and spread it did. i know this because i got email notifications from twitter when other people's tweetdeck clients automatically retweeted the tweet that that my client automatically retweeted. that's actually one of the things i think i did right - i have twitter setup to send me notifications for as much of that kind of activity as i possibly can. the result is that i get what is essentially an activity log sent to my email in near real-time and that alerted me to the problem within minutes of it occurring.

that quick notification allowed me to undo the retweet before it propagated from my account again. that limited the extent to which i contributed to the spread of the worm. acting quickly to neutralize the threat in my twitter stream is another thing i believe i did right.

unfortunately i also did a number of things wrong. for example, i knew about the XSS vulnerability before i encountered the worm, i saw excellent preventative advice and even retweeted that, but i failed to follow it exactly. the advice was to sign out of tweetdeck and then de-authorize the app in twitter. what i did instead was close the tweetdeck tab in my browser and de-authorize the app. i took a shortcut because i didn't believe anyone i followed would actually tweet anything malicious. i didn't anticipate that they might do so involuntarily - the possibility of something like the samy worm from years past never occurred to me. and so when news spread that the vulnerability had been fix and that users needed to log out and back in again to apply the fix i re-opened the tab, re-authorized the app (because that was the first prompt i was presented with) and then went hunting for the logout button. that's when i got the email notification that another user had retweeted one of my retweets.

however, i did not see the alert popup that was supposed to indicate the worm had executed. i didn't realize it at the time but that was important because it meant there was more going on than i realized. it meant that the worm had not executed in the client i was sitting in front of. what i had forgotten was that i had another tweetdeck client open on a computer at work and when i re-authorized the app the worm executed on the work computer rather than my home computer. it wasn't until i was on a bus to see an old friend that the significance of what had (and had not) happened clicked and then it wasn't for another several hours before i could get access to that work computer (where the alert popup was still patiently waiting for me) in order to log out and back into tweetdeck again, which i did without de-authorizing the app beforehand so the un-retweeted tweet got re-retweeted.

in short it was a comedy of errors.

what i've taken away from this is a number of things:

1. i am once again humbled by the clear demonstration that i am not perfect. while i certainly knew conceptually that i wasn't perfect, i have had a surprisingly good track record with malware. having my ass handed to me made the appreciation of my imperfection much more visceral.
2. i've gained a better appreciation for the value of de-authorizing apps in twitter. to a certain extent it can seem kind of abstract but what it's actually doing is isolating a vulnerable component from the rest of the network not unlike pulling the network cable out of an infected computer did back when worms that enumerated network shares or sent mass emails were prevalent.
3. i've identified my failure to log out of things (not just tweetdeck but all sorts of sites) as a bad habit. it's pure laziness and it's not even rational laziness because there's almost no effort involved in logging in when you use a password manager. part of the reason i didn't post this sooner is because i wanted to see if breaking this habit was a reasonable expectation or whether saying i was going to improve was just wishful thinking. so far this improvement seems like an entirely reasonable expectation - i've had no problems logging out of things when i don't need the session open any longer.

at the end of the day, improvement is what sets an incident apart from a failure. the only real failure is a failure to learn from your mistakes and do better the next time. i'm not perfect (no one is) but each time i screw up i make sure i get better.

symantec anti-virus is dead

there's a lot of digital ink getting spilled right now over symantec's brian dye saying that anti-virus is dead (one, two, three, four, five, and more to come i'm sure), but i don't see many people asking the tough question, which is "why should we believe symantec now"?

looking back over my past posts about symantec paints a pretty unappealing picture, and reveals what might be considered a pattern. virtually right from the beginning they named their consumer anti-virus product after a man who famously said computer viruses were an urban legend. then, when they then tried to reinvent themselves with their "security 2.0" campaign, they claimed the virus problem was solved. now, when it appears they're trying to reinvent themselves again, they're saying that anti-virus is dead. it seems that whenever their business plan calls for serious marketing, they latch on to messages that grab attention but whose reality is questionable at best.

when the biggest anti-virus vendor starts saying anti-virus is dead, there's no way that isn't going to grab a lot of attention. it seems designed to hurt the very industry they're on top of, while they are (apparently) in the process of trying to distance themselves from it. i've noted in the past that the biggest players in the industry are hurt the least by the consequences of their bad acts. as market leaders they control perception not just of themselves but of the entire industry, so that even if a smaller player wanted to try to present a more reasonable and accurate view of things in order to better compete on technical merit rather than deceptive marketing manipulation, there's very little impact they could have. saying that anti-virus is dead while simultaneously trying to position themselves as something else is essentially a scorched earth tactic.it will hurt the entire anti-virus industry while drawing attention to the alternate industry they're trying to create/break into.

when the biggest anti-virus vendor starts saying anti-virus is dead, there's also no way that shouldn't raise the hairs on the back of your neck. out of the blue symantec starts mimicking exactly the same message that enterprise level infosec people have been saying for years? am i the only one who thinks that sounds like it belongs in the too good to be true category? this is the same kind of technique a malware writer might use to trick you into trying out his/her handiwork. before you get any ideas about symantec using 'trojan marketing', though, it's also the same kind of technique AV marketers used when they told people just using AV would solve their security problems. too good to be true has been part of the AV marketing arsenal from the very beginning, it's just that this new one about AV being dead seems to be designed for a much more select class of dupe, i mean user. this is the same shit, it's just a different pile.

it'll probably work, though. telling people what they want to hear is unfortunately quite effective. even smart people will fall for it, because despite being smart, those people still want to hear something that is far too simplistic to have anything in common with reality. when you look closely enough, the truth always seems to wind up being messy and complicated, not something that could fit in a sound-bite.

this is the reason why i try to convince people to stop listening to marketing (and really, everything that comes out of a vendor is marketing to some degree). this is almost certainly nothing more than another in a long line of efforts to deceive and manipulate the market. if you must listen to something, listen to their actions. they aren't retiring their AV product, so how dead can AV really be?

all that being said, i actually do welcome their shift in focus from purely prevention to now include more detection and recovery. it's about time AV vendors started getting serious about the last 2 parts of the PDR triad (prevention, detection, recovery). it doesn't have to be purely service-based detection though. years ago we had generic detection tools (such as integrity checkers) that end users could use themselves. symantec's focus on providing detection services instead of detection tools belies a philosophy of not trusting the users' competence, which in turn is consistent with their long history of failing to educate, elevate, and empower their users. maybe that kind of paternalism is appropriate for home users, but enterprise security operations? i thought we could expect enterprise level IT and infosec professionals to develop skills and expertise in these kinds of areas, so why is symantec choosing a path that takes these things out of advanced customers' hands?

as much as it seems like symantec is doing an about-face, they really haven't changed their tune. telling enterprises what they want to hear is just a ploy so that enterprises will get in bed with them (that's just what we call pillow talk, baby). they still aren't giving their users any new power to affect their own security outcomes. so far they're just offering words. nothing but sweet, sweet words that turn into bitter orange wax in your ears.

goto fail refactored

i wrote the lion's share of this a while ago but wasn't sure i wanted to publish yet another post about GOTO here since this isn't a programming blog. my mind was made up yesterday when i read this post by Steven J Vaughan-Nichols where he quotes a number of technology personalities essentially giving bullshit excuses for why GOTO is OK to use. it's no wonder 2 separate crypto libraries (both making prodigious use of GOTO) suffered embarrassing and dangerous defects recently when programming thought leaders perpetuate myths about structured programming.

i'm providing this as an object lesson in how to avoid the use of GOTO, especially in security-related code where a higher standard of quality is sorely needed. i'll be using Apple's Goto Fail bug as the example. here is the complete function where the fail was found, with the bug intact:

static OSStatus SSLVerifySignedServerKeyExchange(SSLContext *ctx, bool isRsa, SSLBuffer signedParams, uint8_t *signature, UInt16 signatureLen)
{
    OSStatus        err;
    SSLBuffer       hashOut, hashCtx, clientRandom, serverRandom;
    uint8_t         hashes[SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN];
    SSLBuffer       signedHashes;
    uint8_t         *dataToSign;
    size_t          dataToSignLen;

    signedHashes.data = 0;
    hashCtx.data = 0;

    clientRandom.data = ctx->clientRandom;
    clientRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;
    serverRandom.data = ctx->serverRandom;
    serverRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;


    if(isRsa) {
        /* skip this if signing with DSA */
        dataToSign = hashes;
        dataToSignLen = SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN;
        hashOut.data = hashes;
        hashOut.length = SSL_MD5_DIGEST_LEN;
        
        if ((err = ReadyHash(&SSLHashMD5, &hashCtx)) != 0)
            goto fail;
        if ((err = SSLHashMD5.update(&hashCtx, &clientRandom)) != 0)
            goto fail;
        if ((err = SSLHashMD5.update(&hashCtx, &serverRandom)) != 0)
            goto fail;
        if ((err = SSLHashMD5.update(&hashCtx, &signedParams)) != 0)
            goto fail;
        if ((err = SSLHashMD5.final(&hashCtx, &hashOut)) != 0)
            goto fail;
    }
    else {
        /* DSA, ECDSA - just use the SHA1 hash */
        dataToSign = &hashes[SSL_MD5_DIGEST_LEN];
        dataToSignLen = SSL_SHA1_DIGEST_LEN;
    }

    hashOut.data = hashes + SSL_MD5_DIGEST_LEN;
    hashOut.length = SSL_SHA1_DIGEST_LEN;
    if ((err = SSLFreeBuffer(&hashCtx)) != 0)
        goto fail;

    if ((err = ReadyHash(&SSLHashSHA1, &hashCtx)) != 0)
        goto fail;
    if ((err = SSLHashSHA1.update(&hashCtx, &clientRandom)) != 0)
        goto fail;
    if ((err = SSLHashSHA1.update(&hashCtx, &serverRandom)) != 0)
        goto fail;
    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
        goto fail;
        goto fail;
    if ((err = SSLHashSHA1.final(&hashCtx, &hashOut)) != 0)
        goto fail;

    err = sslRawVerify(ctx,
                       ctx->peerPubKey,
                       dataToSign,                /* plaintext */
                       dataToSignLen,            /* plaintext length */
                       signature,
                       signatureLen);
    if(err) {
        sslErrorLog("SSLDecodeSignedServerKeyExchange: sslRawVerify "
                    "returned %d\n", (int)err);
        goto fail;
    }

fail:
    SSLFreeBuffer(&signedHashes);
    SSLFreeBuffer(&hashCtx);
    return err;

} 

one of the things you might notice is that all roads lead to "fail:", meaning "fail:" isn't really just for failures, it's for clean-up.

another thing you might notice is that the final "goto fail;" doesn't actually bypass any code - it's completely redundant and if it weren't there the next thing to execute would still be the code after the "fail:" label.

the first thing we're going to try is the most obvious approach to refactoring this function, to get rid of GOTO by making proper use of IF.

static OSStatus SSLVerifySignedServerKeyExchange(SSLContext *ctx, bool isRsa, SSLBuffer signedParams, uint8_t *signature, UInt16 signatureLen)
{
    OSStatus        err;
    SSLBuffer       hashOut, hashCtx, clientRandom, serverRandom;
    uint8_t         hashes[SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN];
    SSLBuffer       signedHashes;
    uint8_t         *dataToSign;
    size_t          dataToSignLen;

    signedHashes.data = 0;
    hashCtx.data = 0;

    clientRandom.data = ctx->clientRandom;
    clientRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;
    serverRandom.data = ctx->serverRandom;
    serverRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;


    if(isRsa) {
        /* skip this if signing with DSA */
        dataToSign = hashes;
        dataToSignLen = SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN;
        hashOut.data = hashes;
        hashOut.length = SSL_MD5_DIGEST_LEN;
        
        if ((err = ReadyHash(&SSLHashMD5, &hashCtx)) == 0) {    
            if ((err = SSLHashMD5.update(&hashCtx, &clientRandom)) == 0) {    
                if ((err = SSLHashMD5.update(&hashCtx, &serverRandom)) == 0) {    
                    if ((err = SSLHashMD5.update(&hashCtx, &signedParams)) == 0) {    
                        if ((err = SSLHashMD5.final(&hashCtx, &hashOut)) == 0) {    
                            hashOut.data = hashes + SSL_MD5_DIGEST_LEN;
                            hashOut.length = SSL_SHA1_DIGEST_LEN;
                            if ((err = SSLFreeBuffer(&hashCtx)) == 0) {    
                                if ((err = ReadyHash(&SSLHashSHA1, &hashCtx)) == 0) {    
                                    if ((err = SSLHashSHA1.update(&hashCtx, &clientRandom)) == 0) {    
                                        if ((err = SSLHashSHA1.update(&hashCtx, &serverRandom)) == 0) {    
                                            if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) == 0) {    
                                                if ((err = SSLHashSHA1.final(&hashCtx, &hashOut)) == 0)    {    
                                                    err = sslRawVerify(ctx,
                                                                       ctx->peerPubKey,
                                                                       dataToSign,                /* plaintext */
                                                                       dataToSignLen,            /* plaintext length */
                                                                       signature,
                                                                       signatureLen);
                                                    if(err) {
                                                        sslErrorLog("SSLDecodeSignedServerKeyExchange: sslRawVerify "
                                                                    "returned %d\n", (int)err);
                                                    }
                                                }
                                            }
                                        }
                                    }
                                }
                            }
                        }
                    }
                }
            }
        }
    }
    else {
        /* DSA, ECDSA - just use the SHA1 hash */
        dataToSign = &hashes[SSL_MD5_DIGEST_LEN];
        dataToSignLen = SSL_SHA1_DIGEST_LEN;
        hashOut.data = hashes + SSL_MD5_DIGEST_LEN;
        hashOut.length = SSL_SHA1_DIGEST_LEN;
        if ((err = SSLFreeBuffer(&hashCtx)) == 0) {    
            if ((err = ReadyHash(&SSLHashSHA1, &hashCtx)) == 0) {    
                if ((err = SSLHashSHA1.update(&hashCtx, &clientRandom)) == 0) {    
                    if ((err = SSLHashSHA1.update(&hashCtx, &serverRandom)) == 0) {    
                        if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) == 0) {    
                            if ((err = SSLHashSHA1.final(&hashCtx, &hashOut)) == 0) {    
                                err = sslRawVerify(ctx,
                                                   ctx->peerPubKey,
                                                   dataToSign,                /* plaintext */
                                                   dataToSignLen,            /* plaintext length */
                                                   signature,
                                                   signatureLen);
                                if(err) {
                                    sslErrorLog("SSLDecodeSignedServerKeyExchange: sslRawVerify "
                                                "returned %d\n", (int)err);
                                }
                            }
                        }
                    }
                }
            }
        }
    }

    SSLFreeBuffer(&signedHashes);
    SSLFreeBuffer(&hashCtx);
    return err;

}

as you can see this version of the function is quite a bit longer as well as being deeply nested. this is the kind of code that actually makes programmers think the use of GOTO isn't as bad as their teachers told them it was, because that deep nesting makes the function seem more complex and more difficult to read. on top of which there is a considerable amount of duplicated code. neither of these things are appealing to programmers because they make reading and maintaining the code more work.

however, this is the most simple-minded and unimaginative way to refactor the original function. if we were to also tackle that complex pattern used in virtually all of the IF statements at the same time as getting rid of the GOTOs, we would instead get something like this:

static OSStatus SSLVerifySignedServerKeyExchange(SSLContext *ctx, bool isRsa, SSLBuffer signedParams, uint8_t *signature, UInt16 signatureLen)
{
    OSStatus        err;
    SSLBuffer       hashOut, hashCtx, clientRandom, serverRandom;
    uint8_t         hashes[SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN];
    SSLBuffer       signedHashes;
    uint8_t         *dataToSign;
    size_t          dataToSignLen;

    signedHashes.data = 0;
    hashCtx.data = 0;
    err = 0;
    
    clientRandom.data = ctx->clientRandom;
    clientRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;
    serverRandom.data = ctx->serverRandom;
    serverRandom.length = SSL_CLIENT_SRVR_RAND_SIZE;


    if(isRsa) {
        /* skip this if signing with DSA */
        dataToSign = hashes;
        dataToSignLen = SSL_SHA1_DIGEST_LEN + SSL_MD5_DIGEST_LEN;
        hashOut.data = hashes;
        hashOut.length = SSL_MD5_DIGEST_LEN;
        
        err = ReadyHash(&SSLHashMD5, &hashCtx);
        if (err == 0)
            err = SSLHashMD5.update(&hashCtx, &clientRandom);
        if (err == 0)
            err = SSLHashMD5.update(&hashCtx, &serverRandom);
        if (err == 0)
            err = SSLHashMD5.update(&hashCtx, &signedParams);
        if (err == 0)
            err = SSLHashMD5.final(&hashCtx, &hashOut);
    }
    else {
        /* DSA, ECDSA - just use the SHA1 hash */
        dataToSign = &hashes[SSL_MD5_DIGEST_LEN];
        dataToSignLen = SSL_SHA1_DIGEST_LEN;
    }

    if(err == 0) {
        hashOut.data = hashes + SSL_MD5_DIGEST_LEN;
        hashOut.length = SSL_SHA1_DIGEST_LEN;
        err = SSLFreeBuffer(&hashCtx);
    }
    if (err == 0)
        err = ReadyHash(&SSLHashSHA1, &hashCtx);
    if (err == 0)
        err = SSLHashSHA1.update(&hashCtx, &clientRandom);
    if (err == 0)
        err = SSLHashSHA1.update(&hashCtx, &serverRandom);
    if (err == 0)
        err = SSLHashSHA1.update(&hashCtx, &signedParams);
    if (err == 0)
        err = SSLHashSHA1.final(&hashCtx, &hashOut);
    if (err == 0) {
        err = sslRawVerify(ctx,
                       ctx->peerPubKey,
                       dataToSign,                /* plaintext */
                       dataToSignLen,            /* plaintext length */
                       signature,
                       signatureLen);
        if(err) {
            sslErrorLog("SSLDecodeSignedServerKeyExchange: sslRawVerify "
                        "returned %d\n", (int)err);
        }
    }
    
    SSLFreeBuffer(&signedHashes);
    SSLFreeBuffer(&hashCtx);
    return err;

} 

not only does this follow almost exactly the same format as the original function (thereby retaining it's readability), it makes the condition checking simpler and easier to read, and it has virtually the same number of lines of code as the original.

combining assignment and equivalence tests into a single line IF statement was clearly intended to reduce the overall size of the source code but it failed, and in the process it made the code more complex and difficult to read. the combined assignment/condition checking and GOTO statements were complementary to each other. they supported each other and jointly contributed to the complexity of the original function.

this third version of the function, by contrast, has neither complex expressions nor the potential for complex control flow. the only real complaint one might make is that after an error occurs in one of the many steps in the method, the computer still needs to perform the "if(err == 0)" check numerous times. however that is only true if the compiler can't optimize that code, and checking the same variable against the same constant value over and over again seems like the kind of pattern a compiler's optimization routines might be able to detect and do something about.

complexity is the worse enemy of security, sloppiness begets complexity, and GOTO is a crutch for sloppy, undisciplined programmers - it is part of that sloppiness and contributes to that complexity, even when it's supposedly used the right way. what i did above isn't rocket science or magic. the same basic technique can be used in any case where GOTO is used to jump forward in the code (if you're using it to jump backward then god help you). the excuses people trot out for the continued use of GOTO not only make them sound like dumb-asses, it leads to lesser programmers trying to follow their lead and doing much worse at it. it is never used as sparingly as the gurus think it should be, and even their own examples occasionally contain redundant invocations of it, thoughtlessly applied.

if you actually work at adhering to structured programming rather than abandoning it the moment the going gets tough, you will eventually learn ways to make it just as easy as unstructured programming, you'll be a better programmer for having done so, and your programs will be less complex, easier to validate, and ultimately more secure.