Schneier on Security

Clive Robinson • February 27, 2021 11:26 PM

@ xcv,

But it’s not that easy to mine gold. Tons and tons of dirt have to be moved, and even at that, you’re lucky to break even with an ounce or two of the stuff

Which is why there is a lot of Gold in repositories that is never ever going to come out because it’s feared it’s fake and nobody want’s to be caught holding rods/blocks of steal or tungsten wrapped in a little real gold…

What some call “very fine” or just “top of the line” fake gold turns up rather more often in repositories than those trading in the “paper” certificates would care to admit.

Most people who own gold or trade gold never ever see or touch it or in some cases even know where it is. Part of the reason for this is the extrodinary high price it costs to get gold into a repository due to the testing involved.

Once in a well respected repository it’s “assumed” to be real 999.9% fine all the way through… But somebody realised that you could beat the tests, because they were only mass, volume and importantly surface tests.

That is the gold bar should have the right bullion house stamps and serial number to match the certificate, and the right “density”. Those are the easy tests. Then at the surface… match the color, surface hardness, chemical, and nuclear properties of gold within a very very small margin. Of these only the “nuclear properties” had any “depth” in testing and that was a matter of millimeters at most.

Thus the middle of the bar or even coins was a mystery unless you smelted it down and recast it (see gold VAT fraud where coins got melted down into rough bars, some cast in the frog of london bricks…).

Thus taking other elements from the periodic table that did not have magnetic properties[2] you could make an alloy of the same density of gold… In higher quality fakes the main metal used was(/is?) tungsten. In early fakes it was in the form of rods, but these can be “reliably non destructively tested” for with I’m “told” certain resonance techniques these days[3].

But you have to consider something,

1, Who’s going to test their own certified gold?

2, Which bullion house is going to want it known they have in the past certified fake gold?

The reality is only people who are going to use gold in industry, or are going to move gold from one repository to another for some reason and have to re-certify it[1].

Because what is traded mostly is not the gold but the certificates, which are in effect “bullion backed currancy”, and it’s way way easier to test if the certificates are genuine or not than the gold. So it is just another game of “confidence” based on the illusion of testing and auditing.

Which is why I asked you the question you avoided about what capitalism does to hard assets.

And lets be honest, if you have gold bullion in a repository what are you going to do if you do test and find you only have a quater the gold or less you thought you had?

Who’s insurance is going to pay up on that one? Or will they just argue it out in court for ever rather than pay up?

[1] Ethiopian Gold scandle,
http://news.bbc.co.uk/1/hi/7315137.stm

[2] In the Ethiopian case, apparently steel was used as the core material. However you can detect that fairly easily. One trick I’ve known about for five decades came about due to an argument between me and another pupil in a school physics practicle. It is quite simple all you need is four bar magnets, four button compasses aranged in a square, you adjust the position of the magnets to get a null on the compasses anything with magnetic properties that gets put off center in the square causes the needles to deflect slightly, moving the object around in an off center circle makes it fairly obvious even with very tiny deflections. Think of the setup as a primitive magnetic equivalent of an electrical bridge circuit.

[3] The reason I am somewhat cautious about the claimed test of “resonance techniques” is as an engineer I know how to deal with numerous types of resonance effects, the trick is knowing which are being tested for…

Clive Robinson • February 28, 2021 2:38 AM

@ lurker, SpaceLifeForm, All,

PM in press conference admits to being disappointed; her body language suggests a stronger emotion.

It happened because of a political economic decision, so as the person at the top it kind of is her responsability.

What should happen when you test some one is not tell them to “go home and wait”… any teenager could tell you what’s going to happen, or anyone who has seen “Ferris Bueller’s Day Off” from the 80’s.

So what you do is you “quarantine the person” and “pay them for their time” untill the test results come through.

With some testing all you would realy need to do is sit them down with a cup of coffee and a biscuit and get them to fill out some form, the tests are that quick. That first test should be one that errs on the side of false positive, which should be enough to decide if a person needs further testing that takes longer. I mentioned this back early last year some time as it also cuts costs to use graded testing, with a finger prick test sub $10 to do and lab test more than $400 to do, you get the results you want for less cost and over all less inconvenience.

One of the reasons we have remand centers and bail, is that over a hundred years or so of policing we know people will not do as they are asked, so they have to be persuaded in some way.

But on the bright side, the fact they’ve got things back under control should give people more confidence in the track and trace systems on place.

Though I would be tempted to sling the young lady concerned into “hotel quarantine” at her own expense, without a phone, internet, TV, etc, it might make others think a bit…

Mind you the UK government has decided “I’m Special” and thus now an honary 80 year old. Along with all the paperwork telling me I must “sheild” for not just my protection but that of the country is a long list of things I’m not supposed to do…

Like, Do not go to a hospital except in emergancy, Do not go to a Doctor unless called into the surgury, Do not go to the pharmasist to get perscriptions, Do not go shopping even for essentials, Do not go outside for excercise and so on and so on. As I went down the list I was fully expecting to see Do not go to the loo…

But seriously how do they expect you to survive? Well apparently I’m supposed to have a “care bubble” AKA family I live with or “carers”. But I live on my own… So that means it would all have to be done by strangers such as charities and social service workers etc. So I’d end up having contact with more people than I already do, and worse those people will have had contact with many dozens of other people per week. Far more than the average “lock down bound” shopper would…

But it gets better… Apparently I am also entitled to “free priority home delivery” from shops that have such services… All I have to do is register on their databases as being “special”…

Let’s put it this way, I kind of look like a scary stand in for a Klingon Storm Trooper, and capable of walking through brick walls and picking up cars with one hand (admittedly something I’ve not done recently). I’m supposed to convince people I’m the equivalent of an 80+ little old lady… Yeah right, only in a bureaucrats dreams. I’ve spent a year avoiding “the plague” in one of the highest risk areas in the “dirty man of Europe” if not the world at one point hopefully I can still manage to “be safe” by myself without having to tell the world ‘The government says “I’m Special”‘ it would be “Just so embarrassing”.

Clive Robinson • February 28, 2021 9:34 AM

@ Winter,

This is more or less a subplot of Terry Pratchett, “Making Money”.
Great book about money, the illusion.

And very crul and unusual punishments as well 😉

Clive Robinson • March 1, 2021 4:32 PM

@ Joe K

To that end, I’ve spent the morning working about halfway through…

Yes it was a 1st of April RFC but like all such it’s only amusing if close to the truth.

This quote from the RFC,

“The above question arises from the serialization process which is
performed on messages in order to send them through communication media.”

Is as true today as it was back in the 1950’s when computer architectures came in all sorts of flavours.

Part of the problem is we in the west write our words from left to right, but our numbers are actually written from a “decimal point” that appears on the right. This is due in part to how we say the integer parts of numbers.

But when in mathmatics we write them down as arrays holding polynomials A[0] is usually the least significant digit, that way the number powers aline with the index. So xB^i is B[i] holds “x” and the meta-data is the radix or base value so 10 for decimal 16 for hex.

But which way does that array exist in memory is A[0] at the lowest value address or the highest.

What of the actual integers that form the array? If they are 8bit bytes it’s not a problem, but when you get a 16bit word of two bytes then you have two options {0:1}{1:0} which is not too bad but what of a 32bit quad byte? Well depending on the way the hardware works it could reasonably be one of four options {0:1:2:3}{2:3:0:1}{1:0:3:2}{3:2:1:0} though there could be others I’ve not seen them. I’ll let you work out the options for 64bit or 8byte integer numbers.

But it gets worse… With 8 bits you get one byte, two nibbles, or two BCD digits. All of which can be either way around. And yes you can have an ALU that has nibbles one way around and BCD digits the other way around, I actually helped design such a monster and I’ll leave the war stories behind that for another day but in short it was due to I/O and trying to get maximum data transfer whilst having a one size fits all I/O socket…

All fine if the computer system is isolated, works in groups of the same machine type and importantly the majority of the programmers work not at the ISA level but at some higher level language that just hides the horrors under an abstraction layer.

But what happens when you need to communicate with other computers? When I started my design career RS232 and IEEE488 were past their ascendancy and were on their way down in favour of “networking”. There is a lot of nonsense spoken about networking but from a software point of view how bigva block of data bits is is important. What many incorrectly call RS232 is supposed to be seven bits of data a parity bit and a start bit and one or one and a half stop bits… That means much of the time you have ten bits on the line for every seven data bits sent thus 70% –or worse– bandwidth utilization. The start and stop bits are there forvboth electrical and mechanical synchronization. When TTY’s went “glass” and VDU’s started to proliferate, people started to want to squeeze more data down the line in a given time period. Part of the reason was TTY’s were “line oriented” and VDU’s could be either line or “screen oriented”. Screen oriented could be done in one of three ways,

1, Cursor moving.
2, Cursor addressing.
3, Screen refreshing.

The first was horrible the green box would visably move along and up lines like some sick pacman. The second the green box would jump around the screen, it was much faster but required tricks outside of the available ASCII control charecter set. Also there was a probblem where was the origin or char 0,0? Bottom left thus compatable with line oriented programs, or top left, compatible with how we write a page of text… As you can guess some people went one way and others the other. But Top left had a problem when you got to line 24 or what ever the bottom line was numbered you found yourself going from screen down to line up mode[1].

Screen mode was way easier, at it’s most primative you just sent a clear screen command and an entire 80×24 screen of charecters in one go. The down side to this was it was slow. In part becsuse 9600baud was the top speed many could use but that 70% utilisation made things a whole lot worse thus it would take 80x24x10/9600 or two whole seconds to redraw the screen… Getting it to 1.4 secs by taking out all those start and stop bits would be advantageous not just for computer to VDU transfers but also computer to computer transfers via CP/M’s pip command which was sometimes the only way you could port software or print files from one computer to another, especially if it was an 8inch machine to a 5&1/4 machine.

So the char by char “asynchronus” terminal codes got changed to block by block “synchronous” network codes like “High-Level Data Link Control”(HDLC) which importantly was bit not byte orientated… Hence avoiding bit ordering in byte issues. As a synchronous protocol it sufferes from the NRZI problem where long runs of zeros do not send line level translations thus synchronization can be lost. There are two basic solutions use a lower level physical “Manchester Encoding” or split the data into 8bit bytes and send them asynchronously as is seen in early protocols prior to and including PPP still used for dialup data connections to the Internet.

But later network codes got more complicated and 8bit blocks were inefficient to send binary data. Which is where the “network endian” problems arose. And why we have “network order” now as effectively a standard.

But you asked,

If you care to elaborate on this point, I am curious.

With regards to ASN.1.

Well “network order” is realy just an ordered “bag of bytes” which is quite a low level form of “Abstract Data Type”(ADT) and can not even handle 16bit integers. ASN was designed in part as a way to Formally Describe data objects more complex than a byte and how they get serialized.

ASN.1 is very powerfull and very discriptive however there are problems. Two simple ones are,

1, Ranges in specifications are optional.
2, Some oitput formats (using PER) do not contain meta-data.

But there is a more subtle issue it’s a form of logic based language, that can not describe it’s self which has implications going back to the 1930’s and the fun to be had playing with Cantor Diagonals.

Mostly you will never notice but occasionally you can end up with a problem you can not get around even though you think you can.

You get a similar problem unsuprisingly with the declaration of data structures in C which has a messy cludge via pointers. Which if you are a programer you might well have seen and had a Oh shite moment. So you want to define a linked list where a first data type implicitly contains a link to a second data type to do this you need to know the size of the second data type. But on declaring the first data type the second data type is unknown so it can not be done. Unless you declare the second data type first, but you can not because it links to the first data type.

In effect you create a “Turtles all the way down” problem. The solution in C is to use “pointers to structures” but that has problems of it’s own when it comes to integrity and you can do real nasties that a compiler can not catch.

ASN.1 has a similar problems, and unless you are a “hot diggity Java” programmer who gets deep joy from serialization then hopefully you will never come up against the problem.

My standing advice to people on serialization is “don’t” unless you realy know what you are doing, it’s the same as with power tools, you can quickly get your self into a real mess and a world of hurt.

If you must then start small and keep it as simple as possible then make it simpler again. Yes it is possible to send complex data objects in one shot across a comms link but why would you want to? Especially if it’s real world comms where your probability of an error goes up with the length of your data packet. You send a large data packet an one bit gets flipped then you have to repeatedly send the large data packet over and over till you get lucky. You might not see this over and over behaviour because it’s hidden by lower layers in the stack but you will see the drop off of usable bandwidth and increased latency. Which might be a disaster if you are guiding an object more than a few light second or more away (a problem remotely piloted drones can suffer from with multiple hop links).

[1] Thus possibly just one of the reasons the *nix library that –partialy– abstracted this from programmers was called “curses”. Actually curses was not exactly original work, it was based on screen handling in vi via the termcap DB. Ripping it out of vi in the 1980’s and making a general purpose library of it made writing other programs like the game Rougue oh so much easier. By the time this happened I was already a grizzled “serializer” and had my own box of tools written in an abstracted form of assembler that alowed fairly easy movment from one computer architecture to another. Because of this box of tools I got seen as some kind of “guru” and was actually paid a fee to write a book on serial data transfer and serial port hardware. Due to commercial moves involving take overs, the book got dropped and suprisingly I got to keep both the advance and the rights. Somewhere I still have a chunk of it on “punched paper tape”[2] a backup medium that has survived unlike all others such as cassette tapes and 8inch floppies and later 5&1/4 floppies

[2] Long story short I had a second hand KSR type paper roll TTY I used as a printer from my Apple. It also had a paper tape punch and reader. One of the mad things in my life was becoming adept at editing text using a couple of such units back to back and a little box I’d designed, a skill that is more redundant than editing audio on cassette tapes with a razor blade and sticky tape…

Clive Robinson • March 1, 2021 8:08 PM

@ SpaceLifeForm, Bruce, Joe K, ALL,

At the hardware level.

I did warn at the time that started as Intel’s little “Ghost of Xmas past” was going to be

“The Xmas gift that keeps on giving”.

And I gestimated atleast half a decade out of it well we are beyond that 2/3rds mark, and it’s still giving…

Only another 20months to go…

Actually if I was truly honest I would probably have said a decade or more. But by then the number of vulnerable platforms will be I hope quite small, therefore the chance of seeing it reported would likewise be very small.

That’s not to say I do not expect other CPU chip hardware vulnarabilities to come up I do. Because the problem with the Intel style processors is they have way to many little “go faster stripe” cludges built in for specmanship, that now the interaction between the cludges is way to difficult to determin. So yes,

Complexity is not your friend.

Clive Robinson • March 2, 2021 8:14 AM

@ MarkH, Tamara, ALL,

I’ve also seen toner transfer to adjacent sheets of paper, but I think I’ve seen this only from copy machines but not laser printers. I don’t have a gauge of how serious this problem may be.

Toner is a very fine dust that is effectively a plastic. It is deposited on the drum by a high voltage “static” charge, it’s then transfered to the paper by mechanical contact and residual static charge to keep it in place, before moving under the heated wire, that melts it.

Much like sealing wax it makes a very simple mechanical bond to the fibers in paper not chemical or absorption. Which means it is almost as simple to remove from paper as it is to put on and will under many circumstances come of with just time and contact pressure, which can be speeded up much like the removing of chewing gum from cloths.

You can make a “one time” Tee-Shirt transfer by printing onto “baking parchment/paper” you have cut to size and attached to a “carrier sheet” (laser printer feed mechanics don’t like certain types of surfaces as the grip rollers don’t grip). You then put the baking parchment toner side down on your Tee-Shirt thats on a hard flat heat proof surface (I’ve used a wooden dining table). You then apply heat through the parchment with a cloaths iron with the steam off and the toner melts, as the parchment has a very fine surface compared to the cloth and you have a little gravity assist 😉 along with the “elbow grease” the heat softened toner grips the cloth. Carefully pealing away the parchment leaves the toner image on the tee-shirt… However because you did not add other chemicals the image often comes off in the next cloths wash (warning toner loves irons, and sticks to them very easily, so do not get the iron on the toner on the tee-shirt).

I used to use a very similar process to make Printed Circuit Boards at home. The toner sticks quite well to the copper surface so provides a mechanical barrier to the ferric chloride etchant solution. It was a fast way to mskr and test an RF prototype, copper clad board can have msny substrates RT Duroid and Arlon being a lot better thsn FR4 stock fibetglass at VHF and above frequencies. Whilst “photo resist” PCB csn be purchased it’s either on paxoline “paper board” or FR4. Whilst you can buy the “photo resist” in a spray can, you have to spray and dry in a dark room which as Terry Pratchett used to say “is a bit of an embuggerance”.

With the right household chemicals the laser toner can be not just more easily lifted from an ordinary paper surface to the cloth, it can be made to bond much better so the image will survive gentle “hand wash” settings. Though as you might have guessed “Where there is a market a product soon comes along” and you can by special paper for your laser printer to do this now…

Oh laser-jets use a proper ink which is why for highest quality images you should use paper with chalk/flour surface finish for them otherwise the “bleed into the paper fibers” and bloom out. This is thus likely to remain in place longer, however I’ve no clue as to how stable the colour chemicals “pigments” in thr ink are. I know early colour inkjet ink faded rather rapidly.

Oh the story behind the ink-jet printer invention is the usuall electronics bod bored out of his brain “farting around”. He had a hypodermic syringe with an organic ink like (flux) substance in it and noticed that by putting the tip of a hot soldering iron on the steel needle he could get a small droplet to fly out as it “flash boiled” the carrier solvent. Did this poor benighted soul make any money out of this now multi billion dollar ink-jet industry? No. Why not? Because although the use of a heating element was “novel” the inkjet using electromechanical means was pattented by Lord Kelvin back in the 1800’s as part of a telegraph recording device. Siemens were the first to make commercial use of the electro mechanical inkjet in a chart recorder in the 1950’s and it kind of worked like the precision ink pens made by Rotring (that nearly died out[1]). Whilst the heated element inkjet did make it to market, others worked out how to use “pizocrystals” insted which is what most inkjets use these days.

[1] Technically they were “technical drafting pens” and whilst the nibs hung out longer on XY ploters they more or less vanished as CAD happened. I have a full set of original pens and nibs in their cases. However whilst “technical drawing/drafting” is done on computers these days various types of calligraphy and art brought them back from the void, though the range of nibs is reduced,

https://www.cultpens.com/i/q/RT04273/rotring-isograph-technical-drawing-pen

And yes I still use mine from time to time, not for “art” but for technical drawings and charts/map masters.

Clive Robinson • March 2, 2021 7:26 PM

@ JonKnowsNothing, ALL,

Makes you think some, if the air you breath is bad enough to erode granite.

And as I mentioned my CD’s made by Phillips…

It also destroys cellulose and ceritin fibers, so wood paper, cotton/flax paper and wools in rugs etc, all of which have been or are currently being used to store information.

Then there is the issue of the speed of degradation to nitratrd celulose that turns film stock into explosively flamable dust etc…

I still get funny looks from people about “ironing clothes” when they talk of the “fresh ironed smell” I tell them that it’s due to the sulfur being turned into acid and eating both their nose and clothes… Yes it’s true your clothes do last longer if you don’t iron them.

Clive Robinson • March 3, 2021 2:53 AM

@ MarkH,

Any thoughts, Clive?

The impression I got last night that my brain was saying “you are reading this wrong” was that by combining two techniques, a speed up technique evolved where by a lot of tests that were going to produce invalid results would be removed and that the factoring process would become linear…

What makes the paper impenetrable by mortals is it is very information dense. You only have to look at one of the pseudo-code boxes to see things have been squeased in, in such a way clarity does not realy result as you would expect.

On the assumption the preprint paper is valid, I would have to first get clatity to gain insight. My normal way to do this is to turn equations into graphs or processess and study the results and look for key variables or activities such as “inner loop” etc[1].

Currently on the “get clatity” to “gain insight” I’m not getting anywhere, in effect I’m falling before I get to that first fence in this race. That is in what is a messy thread of infomation I can not find a suitable point to grasp and work my way in either direction.

If this was a hoax, then presenting it as such a puzzle would be a way to hide it, but you would have to be a clever puzzler to keep the experts uncertain.

Thus I’m tempted to not revisit it untill others more skilled in such arts have got a firm grip and a direction to go.

[1] Think of it if you will like analysing a complex waveform with the DFT. Theory says the waveform is a sum of sine waves, all you have to do is find them and which ones are key to what you are seeing. Astronomers do this with orbits of objects in the solar system where you can not do a two or three body problem reliably.

Clive Robinson • March 5, 2021 3:01 AM

@ JonKnowsNothing, Bruce, ALL,

On bit flips from out of Space

Yes I’ve mentioned this particular problem before last time I think was when RowHammer was fresh or was it Spector, or both… as per usuall the “Not Invented Here” club said “No not possible” yet here we are, and it looks like MS are fresh joiners of the NIH club as well…

Sorry to sound bitter about it but hey blunt as they are even I need to stretch my claws very occasionally 😉

Reading the article,

1, 0ne legitimate site.
2, thirty two sites posible.
3, fourteen available.

So 14/31 = 45% of the possible names.

And this,

“Over the course of two weeks, Remy’s server received 199,180 connections from 626 unique IP addresses that were trying to contact ntp.windows.com.”

So with 45% of the possibles and 626 unique IP adresses,on those… Potentially 1386 machines over the entire range of bit flipped IP addrrsses…

It does not sound a lot and some may be typos, but it’s still impressive for effectively doing nothing other than register fourteen domain names.

As the researcher noted,

“Remy said the findings are important because they suggest that bitflip-induced domain mismatches occur at a scale that’s higher than many people realized.”

As for the quoted Microsoft response,

“Update 2: The Microsoft representatives didn’t answer my questions, but they did say: “We’re aware of industry-wide social engineering techniques that could be used to direct some customers to a malicious website.””

Huhh “Social Engineering” my ass, they are trying to “blow smoke” and failling miserably.

Not that there is much MS can do other than “sign their time signals” as they do with their software updates. To give some modicum of authentication.

As for the hardware, there’s not a lot you can do… The first attempt was to use “Error Correction” by simple parity checking. As any one familiar with how that was done we get “Parrity Checking” with a ninth parity bit which means the memory is atleast 9/8 = 12.5% increase in price, board real-estate, pin count etc. So not desirable in commodity devices. Plus as it uses XOR then only “odd numbers of bit flips” get caught not “even numbers of bit flips”… There are better check codes like Hamming, but they all require incrrasing the amount of memory, which just increases the surface area where those energetic little photons can flip a bit or two.

Whilst there are radiological sources on the ground, the most likely cause of problems is “Coronal Mass Ejections” or “CME’s” it would be interesting to use the researchers project over a longer time frame to see if it can be used as a crude “Space Weather Telescope” with the IP addresses givong approximate geo-physical locations for “the sensor” computers…

Clive Robinson • March 5, 2021 5:48 PM

@ SpaceLifeForm,

Keep in mind that it does not matter where the bit is flipped, because SNI is NOT secure.

SNI is a gaping privacy wound and could have been resolved another way. One such is multiple IPv6 addresses per Host interface for instance. But that to is every bit as much a gaping privacy wound, as are nearly every other suggestion.

The reason for this is because of where the encryption layer is in the stack. The lower it is placed the more secure and more private communications become, as well as suprisingly to some the more flexible.

Moving the encryption layer up the stack reduces security and privacy, eventually you end up with only message contents security with all the traffic meta-data painfully exposed.

It is this reason the military amongst others use,

1, Link layer security.
2, Routing security.
3, Message security.

Together along with other techniques.

I know that the original argument was that SSL required high power resources thus had to be at least at that level of the stack and again it had to be the way around it is because client computers did not have the computing power. But that was back in the 80’s arguments, around the same time leg warmers, sweat bands, shoulder pads and big fluffy hair does were argued to be attractive… These days we think the opposit but we appear stuck in the 80’s for reasons we should be able to break free of…