Schneier on Security

Clive Robinson • August 11, 2018 3:30 AM

@ Alejandro,

I don’t need to hear anything more about Kaspersky. Do you?

Possibly not but the two issues raised by the event are much broader than aby one organisational entity. That is we’ve seen them before and we will see them again over and over.

The first issue is not technicall in the ICTsec sense, but one that is sociological in nature.

Bug bounties, especially limited scope bug bounties can be problematic at the best of times, due to the managment of “gain expectation”[1]. It’s just one of the reasons I don’t like them very much. Which is why I tend to give my research results without any real gain expectation[2].

But there is a downside to even giving the results away. Some time before Stuxnet I described on this blog how you might go about delivering malware to supposadly “off line” or “air gapped” systems in a “fire and forget” manner along with how to do “headless” control channels for bot nets and the like. I can not say for certain that those who designed Stuxnet used my ideas or not. But there was also the issue of “code signing” that @Nick P and myself had repeatedly warned about that likewise got used. Since then as quite a few of “the usuall suspects” know other new ideas that have been first aired on this blog have likewise been used without acknowledgment @Thoth has suffered this on a product idea he was actively developing. Sometimes the usage was for questionable purposes, thus we have become more circumspect in describing what we research. And based on the more recent behaviour of the likes of the FBI / DoJ in more recent times it is probably only just in time.

So the sociological side is not just “gain expectation” managment but also “risk” managment as well that gives rise to “the chilling of free speech” which is very detrimental to society it’s self not just the parts where people research.

But onwards and upwards as they say to the technology issues. VPNs are problematic for many reasons and will always leak information in some form and at some level, even if it is just that “protected data” is disproportional to some expected measure, thus indicative that some covert channel is in use.

There are two issues that are in effect unsolved with VPNs and for that matter with other privacy systems and even just mobile systems. They are “discovery” and “scope”, that is how individual end communications end points find each other and what should and should not be visable to an end point and how.

Currently discovery systems fall into two broad types “broadcast” and “look up” neither is covert by design thus privacy is at best difficult for one or both of the communicating end points. Solving this problem is still “an active research” issue, but there have been steadily less new ideas published. Which probably means people are starting to realise it’s a way harder problem to solve than they first think, and the “easy answers” are already “mined out”. The reality is there may never be a way to have full privacy on end point discovery, especially if both end points are mobile.

But scope of what is and is not inside a VPN covarage and how it should be reached is actually a network stack software issue combined with unknown choke point issues.

The basic network stack model is the “single gateway pluss local segment” model in most cases. That is you in effect use broadcast to the local network segment and everything else you send to the gateway address and make it another “routing systems” problem.

The problem with VPNs is what is and is not the local segment. That is what may appear local at the IP layer may not be local at a lower level. Whilst this normally does not matter it does when security is implemented by gapping and choke points. This is because a single network segment may well have choke points such as fire walls and unidirectional routing within an organisational entity on a single network segment. Oh and the delights of Network and Port Translation to have many more local “internal machines” than there are external “IP addresses”. Allof which can be further munged up by the use of both IPv4 and IPv6 addressing on a single homed host or worse multihomed host.

Put simply there is no way any VPN network software can figure this out without assistance. So one solution is when a VPN is enabled there is no local network segment and just the “gateway” which is the local end of the VPN pipe. This can be a real security problem, because if access to a local server is required, it has to be available to the distant end of the VPN pipe, which can mean the local server has to be made visable to public networks thus anyone who want’s to attack it…

Larger organisations thus generaly configure the distant end of the VPN to be local to the servers, and have all traffic routed through what is the central hub. It’s far from an efficient way of doing things and adds considerable complexity and other issues such as printers.

However for smaller organisations using VPNs from an Internet supplier it can be a compleat nightmare and can end up with very restricted systems that hurt productivity.

Most people can start to think up general solutions but there are always edge and corner cases poping up such as printers, making such first cut solutions at best partial.

[1] For some reason we don’t even have a recognized term for “gain expectation” other than at the extremes such as “greed”. The IC use MICE but that does not even have common agreement on what the individual letters in the acronym represent.

[2] My gain expectations are, firstly people give me a name check if they use my ideas, secondly they buy our host @Bruce two drinks if they should meet him. The first drink is because even if he likes it or not Bruce deserves recognition for this blog, which in a way makes him a publisher of ideas. The second is for @Bruce to pass on to me should we ever meet, or any suitable deserving cause he might come across in the mean time. As far as I’m aware the closest we’ve ever come to meeting is both being in University College London (UCL) university at the same time some years ago.

Clive Robinson • August 11, 2018 11:41 AM

@ Wael,

… including languages that have ~12.5 Million words in their vocabulary!

Whilst I don’t doubt the approximation, ~12.5 Million words kind of defies comprehension. It is after all a little under 2^24 words, or every possible 5 letter password using the upercase alpha charecters, or five letter telegraph codes that are still used for sending MilSpec encrypted traffic…

A quick “napkin calculation” suggests you would need an A4 page size book with 7800 double sided pages with 80 lines of ten 5 charecter groups.

I somehow doubt any individual could memorise them and “a dictionary meaning” for each unless they were in effect self describing compound words. Such as “turninghandletoopendoor” for what in English would be “knob”.

Clive Robinson • August 11, 2018 2:12 PM

@ Wael,

I don’t use “Clouds”.

Nope they are considered “Third Party Business Records” from what I can decipher out of various reports on US judical decisions. Which if true means they are just an NSL away at most from the donught munching, gut hanging, gun slinging “Officers of the Peace” 🙁

But onto nore chearfull thoughts,

I won’t elaborate more on this, but you’re full of wisdom, and can figure it out.

I guess God alone knows for certain who 😉

That would be German 🙂

They do have that reputation. However the thought arises, in English you can move words around in a sentance and still have the exact same meaning, if you did that in a German word and spelled the individual words correctly, would the resulting compound word be a “spelling mistake”?

Yes it is such thoughts that reveal the darker, danke parts of my brain…

Not sure whether I should talk about it, or just do it then share the “product”.

I used to do the former as part of the “sanity checking process”. But the level of accuity for that these days sujests that there is a degree of “playing dumb” to get details out of people so they can make compeating products. @Thoth found this to his cost so he went from sharing, to building and selling product.

Thus recent history suggests “just do it and ‘sell’ the product” is the way to go. If you look at Amazon’s behaviour you will also see that planing for only a single product run then doing something else is also a sensible strategy because Amazon have become recognised as IP theives 🙁

Clive Robinson • August 12, 2018 3:47 AM

@ Spooky,

… every single one of those chips also contained undocumented instructions and behaviors.

Yes but they were usually predictable, as they were usually caused by removing gates –that did address maping in the instruction decode to RTL sequencer– to save real estate or get space to squeaze in other known instructions. They were not entire high power CPU’s with the ability to commit any kind of malicious intent sight unseen.

Further as can be seen by several current projects on the Internet, you could “decap” them and using a human operable optical microscope actually work out the CCT and thus mapping. So we now know for instance that certain “8bit” CPUs were realy “4Bit” CPUs with a little slight of hand to fake 8Bit behaviour.

The price of course that they were “one legged dog slow”.

That said certain ~1USD microcontroler chips have the same capabilities –and more– than 16Bit PDP11 and early 32Bit Vaxen. With the original Unix 2 source code not just available but ported by some enthusiasts…

Thus we could go back to a four serial port multi-user Command Line Interface OS of reasonable repute if we wanted the real “thin client” experience…

Only trouble is where to get real “Glass TTYs”. Because it would feel daft to use such a system with even an old 33MHz 486SX motherboard two port serial one port parallel IO card and 16colour VGA card and monitor (even though I do just that with 8/16/24/32bit microcontroler development boards…

Clive Robinson • August 12, 2018 11:29 AM

@ CallMeLate…,

If I had 1/2 Pfennig(2) for every time a German gently advised “not allowed” against one of my actual or proposed deeds, then I would have enough to buy a beer.

Ah “non permiso -v- permiso” or “Continentals -v- Brits”. You can apparently blaim Napoleon for the “non, non, finger wag”.

Some say that this attitude difference was why Britain was very late joining the EEC and also why Brexit was inevitable. Apparently neither view point is true. Having lived through “the in” and will probably live through “the out” I view it as the first verse of a very slow rendition of the Hokey-Cokey. As Arny almost said “We’ll be back” one way or another, it has an inevitability of nearly all Federal systems.

The problem is clearing out some very antiquated Franco-Germanic views and practices out of the unelected and undenocratic top of the hierarchy, oh and the rampant fraud from top to bottom.

Speaking of the pfenning, it was another Napolionic bastardisation of LSD (what the British “Pound, Shilling and Pennce” was based on, but from the remains of the Roman Empire). For some reason Napoleon liked tens not twelves even though the latter is way more usefully divisible. I guess it was what he was used to since his revoloutionary forebears replaced the “livre tournois” (“Tourns Pound” livre being derived from the latin for “pound” that gives the L in LSD). Put simply as Napoleon went kicking the arse out of Continental Europe the local currency effectively got the reveloutionary treatment and got decimalized…

Britain keeping LSD the smallest valued coin was the “fathing” at 1/4 of a penny, although I have a few most have never ever heard of them as coinage. All they have ever heard is “Not worth a brass farthing” meaning something of little value as it was a forgery of the lowest value coin. The pfenning likewise has a simillar fate as does the picayne which few Americans will recognise, but gives rise to the word “picayunish” meaning “of little worth”.

Thus you could if you wish say to an American “Your oppinion is picayunish” with little fear they would understand what you mean…

But atleast the picayune is worth
A “bit”, more than “my two cents” 😉

Clive Robinson • August 12, 2018 4:21 PM

@ hmm, echo,

For dropping .zfs support? Care to explain that gem?

I can think of a number of reasons, but a “too much testosterone in the room” one is “Not Invented Here” (NIH) syndrom. It’s an “in-group -v- out-group” behaviour that has solid roots all the way down to the primative instinctive tribalism we have in our “monkey brain”…

ZFS was designed by Sun Microsystems and is now a registered as a trademark of Oracle Corporation. Who let’s be honest are not exactly popular in a multitude of circles for a whole heap of self inflicted reasons.

In the stricter FOSS movment the use of even minutely encumbered software is a cause of raised blood preasure, ire and flames in various forms.

But what of the commercial sector? Oracle have a reputation of going over board on what they see as potential Intellectual Property infringment, especially when they have purchased it by buying it up from a take over etc. Oracle -v- Google was fairly public and eye wateringly expensive to both parties and achived little of worth.

Even lawyers get twitchy when faced with that sort of “berserker behaviour”, and advise “long barge pole” treatment of anything such a corporate might think it has as IP. Lest it acts as a magnet or lightning conductor.

Whilst ZFS is a reasonably good file system it comes with a lot of extras, that would need to be supported. If it is also not used by a significant percentage of customers it could easily be seen by the likes of “marketing” and other reputedly “testosterone ridden primates” in walnut corridor as costing more than it is worth to the business.

The thing is men talk about other men as being testosterone riddled and nobody generally blinks an eye or they even laugh amongst themselves about it with expressions about other men having “brass ones”, “more balls than brains” etc. Women likewise make simillar hormon related comments about other women. Nobody sees it as sexist untill it’s said across the sex divide either way.

In essence it’s a basic example of “in-group -v- out-group” or if you prefere a form of tribalism. We will take way more from within our group than we will from outside the group. We see external critisism and we “close ranks” or “circle the waggons” and attack back as a basic response. In efect “killing the messenger” rather than dealing with the message in a rational manner.

Clive Robinson • August 13, 2018 5:07 AM

@ echo,

The first is asteroid impact versus volcanic action causing extinction of the dinosaurs. The second is a look at the unique brain of the elephant.

They almost combine as well…

The elephant is the mountain of land creatures, whilst it does have preditors, untill recently they were only a minor issue. Thus the elephant could aford to think contemplatively.

However if you have ever seen the damage a single “rouge” elephant can do to a village or when a heard decide to remove trees etc you might consider them like volcanoes.

Humans on the other hand are descended not just from the trees but also tree dweling species who’s fight or flight instinct is mostly a case of unreasoned head for the trees, then when relatively safer think in an initial reactive way then occasionaly in a contemplative way to deal with the problem.

The result of such thinking is generally “group action” that is they form groups not just for increased safety as many herbivours do, but like other primates they take it forward into “group hostile action”. In fact more so than most, which has ment that mechanical specialisation that many creatures evolve, is less important than cerebral specialisation.

But this in of it’s self has caused significant issues, mechanical specialisation tends to be a linear outcome process and quickly reaches limits (see peacock tails). The results of cerebral development have tended not to linear increases in outcome but outcomes to an above unity positive power. Which can as we have seen with the Internet result in “An army of one” globe spanning result. Which could have a similar effect to a major asteroid strike.

That is in effect there is no warning that is of consequence and the outcome is wide area devistation if not global weather effecting for decades if not centuries. However we believe that there are larger earth shattering outcomes possible, which accounts for why we have a moon… Man may not yet quite be capable of such destruction levels, but we have certainly thought out how they might be achived not just in fiction but scientific fact. Scarily it turns out to be a time / energy trade. That is little effort and then wait a long time, or big effort and wait a relatively short time. It is these sorts of trade off that usually get mankind into the most trouble…

After all it only takes one idiot with sufficient resources to think “I’m going to die anyway, so I might as well take as many of you as I can with me”. Which as recent years has shown is becoming a more popular idea…

Arguably it’s already happened with the sea state change after 9/11. You could hope it stayes “Security Thearter”… but there are now way to many rice bowls held by short term graspers, thus we have passed several tipping points and the tail spin spiral is getting both steeper and faster… We could already have started to an “extinction event” the question is can we stop it at say the extinction of society, or will it run on to the extinction of mankind, or worse?

It’s one of the downsides of contemplative thinking, you can with some experience see where things can go and actually most outcomes are likely to be detrimental one way or another…

Clive Robinson • August 13, 2018 11:27 AM

@ echo,

The Atlantic article contains my favourit “faux fallacy”,

The gambler’s fallacy makes us absolutely certain that, if a coin has landed heads up five times in a row, it’s more likely to land tails up the sixth time. In fact, the odds are still 50-50.

It’s false because of an unrealistic assumption of “a fair coin” and even more fun, the assumption that “a coin no matter how fair can not land on it’s edge”. I’ve seen the latter twice in my life but the former is unexcusable because how would you test for a fair coin in the first place. Look at it this way how many heads in succession would you need to be confident the coin was not fair?

The answer is there is not any number of all heads all tails or the occasional edge that will give a true as opposed to probabalistic answer…

As for the film you are watching it gets a big ahhh in my house, because of a certain female actress, who after prommoting the film in England decided that staying was a good idea…

Clive Robinson • August 14, 2018 3:20 PM

@ Bob Paddock,

What is a ‘Fair Coin’ if we are in a simulation?

The answer is “whatever the entity running the simulation wants it to be.

Which kind of shoots down Mr Alexander’s argument.

I could go into it at depth, but lets just say any “encompasing” simulation sophisticated enough to alow you to build a “test” simulation within it would be able to not just recognize it as a movable object but more importantly change the inputs to your test simulation as befiting the environment it is in within the encompassing simulation.

Look at it this way it would no the differencr between “it’s raining and you are indoors” and “it’s raining and you are outdoors”. Thus why should the encompassing simulation treat Mr Alexander’s test simulation within the encompassing simulation any differently?

The only way Mr Alexander’s idea would work is if the encompassing simulation was deficient in some way and more importantly being a part of the encompassing simulation you were somehow able to not only become aware of the deficiency but build a test simulation to demonstrate it without the entity running the encompassing simulation beong aware of it.

If you look back through this blog you will find that @Wael and myself have had “turtles all the way down” conversations about the supposed origin of our universe (oh and about Seth Lloyed, who thinks the universe has all the required properties not just of a computer but a program running on one…).

So moving on to,

Are there any other tests better than this to check a random number/event generator?

Definitely for a whole heap of reasons. NIST SP-800-22 “marks a point in time” at which certain statistical tests had been “suggested, discussed, attested, recomended approved and finaly published. By which time the world had moved on and both new questions and tests had arisen.

It’s down to the consequence of times arrow and “known knowns, unknown knowns, and unknown unknowns”…

SP-800-22 Can be looked at as a starting point. Thus if your generator passes it then “It might, only might be a usefull generator”. The chances are that even if it your generator does pass it will still be a fairly ropey generator.

Oh and in the case of a TRNG / TRBG as they are “unbounded” they will at some point fail the tests occassionaly and pass on other occasions.

Clive Robinson • August 15, 2018 8:16 AM

@ Alyer Babtu,

Of course von Neumann also said anyone who tried to arithmetically simulate randomness must be in a state of sin.

He was correct from the theoretical point of view and the lamentable algorithms of the time.

However the world moves on. Consider three words,

1, Random.
2, Chaotic
3, Determanistic.

Can you find sensible theoretical discriptions for them without to many assumptions? That are also meaningful in a practical way that might be of use?

The thing is when you start considering these words from the practical “black box” approach you discover things are not what you might expect. Especially if you are only observing the outputs.

Clive Robinson • August 15, 2018 3:43 PM

@ Alyer Babtu,

Ahh the coin flipping paper… I think it’s been on this blog, but can not remember how far back…

Oh by the way, I can alter the way a “flip up and land in the hand” coin flip comes out, every time 😉

Even after the coin has come to rest and I like every one else can not see it.

I’ve described how before and one commenter said at the time “remind me never to gamble with you”.

Oh just remember I don’t gamble, so… if I offer to make a bet with you it’s fairly certain I already know what the outcome is and I’m doing it to teach you a lesson (I never bet more than a pint so the lesson is cheap 😉

Clive Robinson • August 16, 2018 2:01 AM

@ Thoth, All,

I believe one of the final nails in the coffin for Intel SGX et. al.

For the architectute it’s certainly “dead man walking” but we went through the whys of that back when discussing C-v-P.

There is an old story about “The Russian Bear” from tzarist times. Put simply it was the idea you could not kill the Russian Bear and by implication the tzarist monarchy. Well we all know how that went with the Potekmpin Bridge and later the fall of the Communist replacment.

Intel’s chief exec and presumably other C coridor primates there act and behave like the Intel bear equivalent also has the same longevity and freedom from censure…

Well the end I think has already been written, the question is what can be salvaged. The Bear died but Mother Russia survived, what I wonderwill be Mother Intel, if there can be one…

Clive Robinson • August 17, 2018 3:03 AM

@ VinnyG,

… because nothing current would run in 256 kB (with no paging) even if you could port/translate it.

It depends on what you mean by “current” dare I mention some stack bassed languages such as Forth, that have the advantage of being their own OS if you want and not actually requiring an MMU to do multitasking (unlike most other languages + OSs). As far as I’m aware there is bootloader code for various CPUs that contains Forth you can “rip out” Free Gratis.

As for hardware,

Surprisingly, it looks like 5-1/4″ floppy disks are still readily available…

Not all 5.25 floppies have the same interface as back last century or so due to amongst other things electrical voltage changes as TTL level interfaces became CMOS-TTL then CMOS levels (ie the gap band between TTL high/low and CMOS high/low is different) so you might need to add compatability via resistors etc.

My solution to these problems is generally go RS232 +-5V or +-12V via thos neat little converter chips and if there is no native serial support available on the CPU board just “Bit Bang” a parallel status line or two via a fast timer interupt loop. This has the advantage of giving you access to the newer high tech stuff without real probs (except speed). but the major pluss point for me is RS232 makes a good security choke point that’s easy to instrument, so you can look for nasties or build data diodes with protocol firewalls etc (look up SLIP or PPP if you want Internet protocols but the SLIP stack is way way easier for a whole heap of reasons).

Clive Robinson • August 17, 2018 7:16 AM

@ bttb,

<

blockquote>Does Putin have more respect for Snowden…

<

blockquote>

To be honest I don’t think Putin cares that much as long as they are all pissing out of his boat.

Which is kind of the point realy, unlike US politicians that want everyone else to do and say as they want…

Clive Robinson • August 18, 2018 10:36 AM

@ echo, Alyer Babtu,

Quantum Neural blockchain AI

And similar such random conectletts of jargon, used to inflate the bubble of sunk cost investor wealth still further 😉

This might amuse,

https://www.bloomberg.com/view/articles/2018-08-16/smart-machines-won-t-be-ready-to-do-complex-jobs-anytime-soon

Clive Robinson • August 19, 2018 5:06 AM

@ Alyer Babtu,

Faggin also said something new was needed because a “conventional” computer the size of your typical brain…

Various people have said similar for a very long time. I think even Alan Turing made a similar comment, when the switching device was a thermionic valve (tube) drawing around ten watts in heater power alone.

The point is that the brain does not use hard switching devices like valves, transistors, and the logic gates they are formed into. In essence it uses something like chemical batteries that pump sodium from battery to battery very slowely, and in a form of pulses. Somebody only half jokingly said it worked like a water clock made of lard.

The real arguments though are not about “hardware” because the assumption is we can solve that in some way as it already exists as six pounds of mixed fats.

No the arguments are much more difficult to even think about and are,

1, What is “original thought”
2, Where do “unique ideas originate”.

Oh and the lesser but potentially much harder to solve “how do we recognise/select” them to bring them into conscious use.

Much as Alan Turing insisted that the first computer have a “noise source” in it, later generations have moved on to other more curious ideas of nondetermanistic sources such as quantum effects. The UK mathmetician Roger Penrose upset more than a few people with his ideas, but we are now finding quantum effects poping up in various biological systems such as chlorophyll and smell etc. Almost like a “can’t see the trees for the wood” problem. That is you can see a wood or forrest clearly as a single entity from a distance. But get dropped in without knowing what a tree is, then you have no chance of telling you are in a wood or forrest, let alone recognise it as a single entity.

But after all the fun and games it still gets back to “original” does it actually require “random” would “chaotic” be a better fit[1]… But then how do you go from noise to signal and how do you enhance the signal till it’s usefull.

Simple maths and logic says that the entire process is not a random one amplified to the n’th degree. There has to be some form of direction such as a set of feedback systems tied in with recognizer constructs like matched filters.

But that just moves the problem to that of the filter constants… Which when we get a handle on it, will probably like an air bubble under newly put up wallpaper, just move somewhere else when you bare down upon it.

The result is we are still not anywhere close to “original” or even finding an origin to start looking.

[1] Which in part was what my question to you the other day was about.