Schneier on Security

Clive Robinson • July 7, 2017 7:41 AM

@ ab praeceptis,

I’m under the impression that keys (as in “door lock” not as in “crypto”) aren’t seen and understood properly.

Mostly they are viewed incorrectly. The first mistake people make is not realising they are in general a deterrent only. That is the only real security value they have is as a delaying mechanism, generaly of quite short order. If you see a newtonian “bump key” in use by a trained person, they can aproach and open the door faster than the legitimate key owner can fish the key out of their pocket. Likewise those well practiced with pick guns can open a lock with little time.

Although I have not done it for a while I could “impression” a five lever and ward mortice or five pin cylinder lock and hand file a new master in a lot less than an hour, and could pick many locks in a matter of minutes. Many do not realise that a lot of padlocks and electrical switch –ignition etc– locks are designed badly thus a bent nail or L shaped piece of metal will open them just like the fancy looking key.

As was once observed “Locks only keep the honest honest”…

@ All,

Which brings on a little known fact but is bleedingly obvious once you know. For each locking device be it ward / lever / pin there are very few cutting hights often as few as four. Likewise there may be as few as five levers or pins, and in many lever based locks as they are designed to be opened from both sides all but the middle lever are in pairs hight wise.

Thus there are as few as 12 lever key shapes and maybe 20 pinning patterns…

What makes the difference is “plate depth”, “spindle diamater” and ward gate with lever and ward locks. With basic cylinder locks it’s the cuts / folds in the key blank and pin offsets.

Which gives you the bad news on just how few numbers of commodity locks there are. Thus it would be pointless keeping a database to try to detect unauthorised people getting keys cut.

Oh and one thing to consider if you take a photo of a cylinder lock key when it’s flat on a table to see the pin hight cuts, the cuts and folds in the key blank are not obvious. Likewise when you take an end of key shot to get the cuts and folds of the blank or photo of the lock face plate to do the same thing the pin hight cuts are not obvious. To put it another way, with a little photoshopping there is no connection between the blank cuts and folds and the pin hight cuts. Thus it’s the same issue as having a brief case with two three wheel combination (ie 2 x 1000 possibilities) and a central four or more wheel lock ( thus a six wheel would have 1 x 1000000 possibilities).

As a final thought on cylinder lock blanks, they are often stamped out to give the folds, but the cuts which are more expensive are cut with a horizontal mill wheel. It is very easy to get brass strip and a home milling machine to cut the strip down to the basic box shape. Then using a mill cut out a sufficiently good approximation to the cuts and folds.

Even in very complex supposed “high security” locks the security is more illusory than actual.

In part this is because of the problems with mechanical devices wearing and binding, thus looser tolerances or “slop” is an absolute requirment of the design… It’s why all mechanical locks can be picked or even impressioned.

But also often they are a pointless buy because the door and frame design are way easier to cut through with a petrol driven cutting disk, that bailiffs and court enforcers carry as standard kit.

If you actually want to increase your security a bit cheaply one way is to make a cylinder lock “bind against the door frame” as it makes traditional lock picking and even pick gun usage much harder. One way to do this is have a mortice lock slightly out of line and rubber draft excluding strip. Thus to lock the mortice you have to pull the door hard against the springyness of the rubber with the door pull. In the process the spring latch of the cylinder lock also binds against the door frame. It’s a mild inconveniance to the legitimate user but it slows others down a lot.

Another way is to have a six or eight point deadbolt system. If you go to Portugal or Spain these are quite common. When set up correctly the likes of even the Police have real trouble opening them. Let’s put it this way I’ve seen commercial safes for cash storage that are less secure.

Also have a look at the various door designs used by the notorious Mexican “tunnel rat” Drug Boss. The guys working for him worked out how to make doors not just battering ram proof but also cutting disk / torch near proof as well. Certainly it would delay the authorities so long he could have a sandwich and beer befor leisurely walking to his hidden escape tunnel and be miles away by the time they authorities got in to find him and his lieutenants gone and the tunnel blocked with similar doors etc.

Clive Robinson • July 8, 2017 12:04 PM

@ Gerard van Vooren,

With regards the story of Tim Jenkin, you make a comment about the NSA etc that would make many “old lags” like myself smile.

But interestingly you go on to say Tim created a system,

He improved the system over the years from manual encrypting / decrypting that took hours to a simple program on counted floppy disks with OTP (in the presentation he showed the BASIC code) that were distributed and the phone, with Londen as the go-between.

Which if you replace “London as the go-between” with Moscow Central and ANC with the Russian IC entity then that part of the story would be the same…

The NSA,for years tried to break the Russian system with no success other than when the Russian’s against all knowledge to not do so, reused a percentage of the KeyMat.

The problem with such a centralised system is of course “traffic analysis” of signals that get seen/intercepted, which manual OTP systems are quite vulnerable to. In that although as an analyst you do not get the message contents you do get message length and frequency of sending which gives a broad indicator of work and a fine indicator of changes to routines etc.

There are good reasons not to trust NSA cipher systems some of which I’ve detailed in the past. But the biggy goes back befor the NSA existed and is down to William F. Friedman, working at the privately financed River bank, and hints can be found in some of the River Bank Publication’s he authored.

Basicaly they quite deliberately put mechanical cipher systems into the hands of the armed forces that they knew had variable key strength, some of which (20-25%) were very weak. These were to be used as field ciphers.

The idea behind it[1] may well due to the dual function the NSA and it’s predecesors had. That is to make strong systems for the US whilst also attacking the ciphers of every other nation (and it would also appear now other US IC and associated entities as well including the likes of NATO). What is not certain is where the idea originated, because what is now the UK MI6 pulled similar tricks on the SOE and other entities prior to and during WWII (look up “Poem Codes”).

Further with the likes Tor and the way it’s been designed not to solve the more important “traffic analysis” issues strongly suggests the weak cipher system design game is still well and truely in play.

[1] The idea is relativly simple to grasp and hinges on who issues the KeyMat to be used and their crypto knowledge. If you assume that the cipher system has a range of key strengths from very weak through to strong enough to be secure for a week-month against all attack methods “You” are aware of, then the week-month strength is fine for a field cipher, where messages only have to remain secure for hours not days or weeks. However you also have to consider what happens when your cipher machine inevitably falls into the hands of the enemy. You know that they will pull it appart to find out the mechanics of how it works, and depending on their sophistication will analyse it for weaknesses to the level they can. However untill relatively recently outside of the US, UK, Russia and one or two other places crypto knowledge was to put it politely primative. Thus an enemy not as crypto sophisticated may well copy and use the design –as indeed happened after WWI and WWII– without being aware of the weaknesses the system has. Thus they will –not knowing about the key weakness– select keys at random, sometimes using strong sometimes weak. Unlike the crypto sustems designing agency, that also issues the KeyMat thus can ensure only the strongest keys are used. The thing is that the weak keys will alow the gathering of information that will allows known or probable plaintext attacks which will in most cases short circuit much of the work involved with breaking strong keys. It is known that during WWII both the UK and later the US were breaking enemy high level ciphers and decrypting the contents often before the enemy recipient for whom the message was intended. Certainly fast enough that even field ciphers would be broken in time for a tactical advantage to occur.