Schneier on Security

Clive Robinson • August 15, 2015 6:18 PM

@ Jacob,

Yeh I thought the Kapersky stuff was of not for the humour side of it Nick P also thought so as well.

When people take credit or free rides off of others, sometimes the price they pay is rather more than they first thought 😉

With the NSA changing Suit B algorithms you have to try and guess why. In the past most would have assumed it was caution based on information from research. However post Snowden trove and the Dual Eliptic Curve stich up of NIST you have to ask is there alternative reasons.

Take 2K RSA, in the open community we doubt that they can directly attack 1K RSA, so as my ancestors would say “I din’y ken t’fash”… However they may have discovered an indirect way, we know there have been issues with SSL crypto… But alternativly they may be trying another finesse to bounce people out of using RSA, as legacy code by and large can not use the larger RSA sizes… Which gives rise to other thoughts, so on and so on and “down the rabbit hole we go, following the sounds of ‘I’m late, I’m late for a very important date'”.

Welcome to the land beyond the looking glass where all is not as it seems.

Clive Robinson • August 17, 2015 1:46 AM

@ IETF clownboat,

If you make a lose distinction between “logical” and “mathmatical” ciphers where traditional block and stream ciphers fall in the “logical” and the likes of RSA and EC fall in the “mathmatical” various differences become apparent.

If you look at various attacks and other weaknesses we appear to have reached a kind of stability point with logical ciphers, but not even close with mathmatical…

Which suggests we are way further up the maturity curve with logical ciphers. Where close in system implementation attacks are the current growth area mainly by time based side channels.

This suggests that until the next major attack class (quantum) becomes realistic our current logical algorithms are sufficient, whereas we are still “doggy paddling in the pool” of mathmatical ciphers.

This view point has it’s dangers, as unlike closed community research academic research tends to be more about making a name for yourself than by satisfying “customer needs”. That is the academic market is badly skewed by the “those who have most are given more, whilst those who have least will lose more” principle. Thus the way to academic immortality is to be the first to open up a new field of endevor, not to bring usability or stability to an existing field of endeavor. This is becoming more and more apparent as “tenure” becomes a thing of the past in academia, as a young “proto-academic” you have a choice, become one of the increasing number chasing fewer posts, or go out into the comercial world where research is at best closed, where you will make money but not have either fame or immortality. But more importantly closed work and research holds the community at large in a state of paucity of knowledge thus lack of progress, or put more easily “each man will have to invent his own wheel” which we have seen with the chaotic development of Closed Source Software especially around the likes of the Microsoft Foundation Class (MFC) for Windows.

Such paucity of open knowledge is not a good place to be in with respect to crypto it takes us back to the old pre-DES days, which the likes of the NSA would just love.

Clive Robinson • August 17, 2015 8:02 AM

@ John Gault III,

I’m surprised to not recall seeing any discussion of smart meters as a gateway to backdoors in a variety of appliances, including routers, modems and computers. Call me paranoid, but my working assumption is that new appliances will include all kinds of telemetry, including microphones that will send data through the smart meters.

I must have missed your comment the first time.

There are a number of ways you can look at smart meters as “antisecurity devices”. Discussions in the past have concerned the following,

1, The lack of their secure upgadability.
2, The use of weak protocols.
3, The expected life expectancy of more than 30years.
4, The ability to control home owner equipment.
5, The ability to spy on people via power usage.

If I understand you correctly you are adding another category on to this, that is (6) home owner purchased equipment containing snooping electronics that then use the mains power wiring to exfiltrate data out of the property.

Whilst I’m aware of some entertainment / white goods using WiFi or other network connection to do this like Google’s unit, various TV’s and a very creepy sounding childs toy, I’ve yet to see a reasonably verifiable article on other devices. Though fridge freezers and microwaves and overns have been talked about for years having barcode readers for inventory control and cooking times etc.

The problem for the device is that it needs the user to connect it, to the communications to the internet. That is assuming the user or their neighbors have set their WiFi up securely, the device can not piggyback out to the Internet on it.

The problem with using the mains wiring in the home is the primary limitation of lack of bandwidth and high noise and secondary limitation of no onwards connection to the data collector.

Smart meters thst are designed to communicate over the actual mains wiring are generaly designed to be broadcast receivers, using very very low bandwidth devices on the transmit back haul using protocols not designed for bulk data transportation. This is to prevent overloading of the limited bandwidth. However some smart meters have been designed to use various RF comms systems or other comms including cellular and POTS.

In almost all cases the smart meter network is considered private and thus not generally gatewayed onto the Internet (though this could change). Which would present a difficulty for any other consumer devices doing an ET and phoning home.

So without colusion between the smart meter manufacturer and the white good device manufacturer and the utility provider the current risk appears low.

However as has been observed “there is lots of money in consumer data” and just about any one who can collect it ends up selling it as can be seen by TV shows of Police traffic cop etc activities. So it may happen in the future if the primary limitations can be overcome, so yes it may be a future worry.

Clive Robinson • August 18, 2015 6:57 AM

@ Thoth,

I always wonder how small you can pack a secure chipset because most secure radio setups I have seen especially the military manpack sets are heavy, bulky and their reception simply sucks especially in jungle and cluttered environment. It is just a curiosity if civilians can build better quality secure radio than military.

The reason they are heavy and bulky is primarily not the radio but what goes around it and the specification not just for maintainability and availability, but survivability in quite harsh environments.

It’s becoming a better known fact that “mil spec” is dying as an industry sector, likwise Private Mobile Radio (PMR) as the reliability of celular mobile radios improve.

The think of note is it’s easy to design “small” if you have no intention of making it “user serviceable” or even “repairable”. Which is why the likes of modern mobile phones can be made so thin.

Further if you make it small and light, you don’t need the robustnest of machined metal cases with heavy internal shock mounts etc.

But mobile phones have another advantage, they work line of sight around the bottom of the microwave band. This generaly won’t work in rugged terain unless the cell site has some kind of “Sky Hook” keeping it in a visable place. Military radio generaly works from the top of the MF band through to the mid UHF band depending on the use it is being put to. This necessitates various types of antenna tuning that are quite bulky when made robustly and can easily be more than a third of the radio volume.

Mobile phones also have another advantage, they use TX power down in the milliwatt range, and only need to have a battery that gives maybe 24hours of RX and an hour or so of TX time. Military radios work from 5-500watts, with HF man packs frequently wirking 25-125watts, with RX and TX times vastly increased to the equivalent of 250 and 20 hours respectivly, this gives batteries that take up upto two thirds of some manpacks volume and eight percent of the weight.

I’ve got radio equipment I’ve designed that covers 0.1-3000MHz RX and 1watt TX into 50ohms, it fits fairly easily into a 1U 19″ rack. The physicaly largest component is the SMPSU in it’s shielded case, the second largest being the CPU board that does the various modulation and demodulation at the 100MHz IQ second IF.

It’s a bit old now, but still serves well for doing radio astronomy and amature radio and some older broadcast work. If I was designing it today I’d go for an IF bandwidth of more than 10MHz and a lot more CPU power for the more complex modulation modes around such as DAB and DTV, and try to get around 10dB more on the RX dynamic range which more modern ADCs give.

Clive Robinson • August 20, 2015 8:57 AM

@ Figureitout, and the usual suspects,

Whilst having a hunt round I found myself on this interesting site,

https://embeddedmicro.com/

It has an interesting FPGA system with IO Shields, and what appears an easier to use than normal HLA of Verilog,

https://embeddedmicro.com/tutorials/lucid

Oh and have a look at their basic CPU tutorial, it certainly looks like it takes a lot of the crap out of RTL to logic,

https://embeddedmicro.com/tutorials/lucid/basic-cpu

Clive Robinson • August 20, 2015 7:24 PM

@ Anura,

That said, while I have no ill will to most of those people who would cheat on their spouses and don’t think they should be publicly shamed

Whilst I agree with the “I have no ill will”, they are cheating.

Now I don’t know how it is where you are but in the UK marriage is a publicaly made commitment usually befor a God, supposadly freely entered into by both parties, with amongst other commitments of,

… to love and cherish… forsaking all others… untill death do you part

Thus if they break those publicaly made promises, why should those breaches of promise not become part of the public record?

Just as they would with divorce proceadings or other formal annulment of marriage.

I was once told by a retired army sargent who had fought in both Africa and India during WWII “Son when you get married, your cock belongs to your wife and don’t you ever forget it”, he had a point.

Another friend in the army once said “Marriage is the prelude to divorce, if you don’t get married you can’t be divorced”, which in a some what odd way of looking at it is true. Back when he said it there was little or no co-habiting laws so there was a significant advantage in not being married if things went wrong. But as others have pointed out, marriage is more likely to end a relationship, and baddly than if not married. In the UK there was evidence to show that co-habiting relationships lasted longer on average than marriage for the same age groups…

Clive Robinson • August 21, 2015 4:43 AM

@ Figureitout,

The problem with one way signalling is it has to be robust to non and out of order delivery issues. Whilst one part of this problem “detection” can be resolved fairly easily with sequence numbers, the “loss of data” can not. Thus you are stuck with accepting missing data or providing some kind of forward error correction to account for data loss.

For instance if you think about a toy tank, you send basic left / right instructions and rely on “visual feedback”. Which works ok untill it goes behind the settee etc. If you just continue sending left right etc the unknown errors quickly build up and it becomes chaotic. Thus you change the commands to go to point A, point B, etc, which gets rid of the error build up, but does not help with correcting errors. One way to do this is send “from point C go to point D” thus if the previous “go from point B to point C” message is lost then the tank can make that one lost message error correction. However, it is actually physically one step back in time from where it should be, so there is a “catch up time” issue to contend with, which the tank may not be able to do and thus things become chaotic again. It’s these “chaotic” reasons why forward error correction is not liked from a reliability issue. But also because each forward correction step adds more data to each message and thus requires much greater band width.

To see why “nod of the head” feedback does not work well, consider the cases of out of order and out of time delivery. In the case of the toy tank let’s assume it beeps for each command it recieves. If you as the controler send a command what do you do, do you resend the command again or do you wait then resend, and if waiting how long do you wait? With instant visual feedback it does not matter because you can adjust what you send to correct corrections. But what happens when the feedback is not instant?

I could go on at length, but it should be clear that the feedback is actually both the more important and more difficult part to get right. Further that it gets exponentially more difficult with finer grained control and faster response times. This is something battle field commanders have found out the hardway thousands of years ago, which is why “superior numbers” and “autonomous behaviour” is liked in troop leaders. Such that the commander gives just an objective and a time to accomplish it, to the leader and the troops go out and do it with a reasonable probability of success.

Now consider how such systems respond when the communications is activly under attack, thus both your messages and error correction mechanism needs to be secure and proof not just against “loss” and “out of order” messages but false messages, replays, bit flipping etc in both the forward control channel and return acknowledge channel.

Then move on to dealing with an inside traitor who is trying to hide covert messages within the legitimate forward or return channels.

It should become clear that there is no ideal solution, in that what you do to improve one asspect is counter productive to another aspect. For instance increasing the bandwidth for legitimate traffic, makes covert traffic insertion much easier. Thus trying to control one channel for covert traffic by “clocking” makes other aspects considerably more difficult, in that it makes a DoS attack much much easier.

Thus you have to do a “SWOT” type analysis on each comms path and make the design choices as appropriate. Thus you might view comms paths as “internal” and “external” where internal Threat is traitors, and external Threat is DoS and adjust your design requirments differently for each.

If you have a look at the original Shannon paper on secrecy systems it’s over fourty pages long, and only dealt with one or two of the issues. Thus there is more than a serious books worth to get to know and understand on the subject of comms.

And as I said each comms channel is different so “hair pulling” is to be expected along the way. Which perhaps answers the age old question as to why “bald men are considered wise men”…

Clive Robinson • August 22, 2015 3:09 AM

@ Nick P, and the usual suspects,

https://github.com/siemens/jailhouse/blob/master/README.md

Clive Robinson • August 24, 2015 5:57 AM

@ Nick P,

I was going to look at the zip file over the weekend, but other more pressing issues arose… any way due back at the Hospital later this PM to have another test or three 🙁

Hopefully I’ll get time to download and read it later this week when I’ve caught up with things.