Schneier on Security

Clive Robinson • March 17, 2012 9:31 AM

@ Jonathan Wilson,

And its known (though declassified materials) that IBM and others built computers and special addons for the NSA that were ahead of what was sold to everyone else (the IBM 7950 for one)

Historicaly it’s actually a bit more interesting than that 😉

Various manufactures including the UK’s ICL wanted to put some of these “addons” in as standard but were “persuaded officialy” not to do so.

A couple of the addons in particular were the “set bit count” and the “parity bit” both of which are very basic functions to data communications. Thus the “policy” was well known in the industry long before any “declasified documents”, because lots of people were asking the manufactures to add them and were being fobbed off with increasingly ridiculously excuses. Likewise but to a lesser extent was hardware for programable linear feedback shift registers (LSFR’s). The silly thing about it was that the high speed hardware needed to do these functions was added as standard, but not in the ALU but in the communications circuits.

Another add on that was known about was a fast lookup table, and this function did end up in the ALU as part of fast maths functions. The difference being the ordinary version was “hard coded” with a ROM to do 8 bit multiplication etc whilst the “confidential” version had fast RAM and thus could be programed to do all sorts of interesting things.

Likewise “fast adders” these are a bit of a black art and are still areas of research. However it is known that certain types lend themselves very well to certain “other addition” functions used in cryptography (think addition within a modulus for turning LFSRs into NLFSRs etc)

Another addon that was known to have originated from Alan Turing was that of what we would now call a “True Random Number Generator” based around “Thermal noise”. Refrence can be seen to it in his papers as being especialy usefull for certain types of mathmatical calculations with suggestions it be added to the design then in progress in Manchester.

As far as I’m aware the only non “confidential” computer to have such functionality built in at the time was ERNIE used by the UK Government for it’s “Premium Bonds” draw.

But the whole “confidential ALU hardware” got to be almost farsicle towards the end. The US Government in particular wanted the addons but not to pay the increasingly large costs of “adding them on” so certain manufactures did the “IBM ‘strapping’ trick” of actually building in the functions as standard and having them enabled/disabled by a wire link ‘strap’ on the CPU PCBs.

Quite a few years ago I was shown the guts of a product (of a well known Super Computer Manufacturer of such “NSA” products) that was being scrapped. It should originaly have been “confidential waste” but ended up in a European University who had finaly decided that the electricity bill and leaking artificial blood plasma was not worth the expense any more. I was shown it as they knew I’d been involved with the design of what were the equivalent of super computers (but for things based around FFT’s). It was not to difficult to spot the “confidential” parts and to actualy map out the basic functioning, and work out what the missing instructions were from the micro code instruction set.

So in many respects the whole “confidential ALU hardware” addon’s was well known to very many in the industry.

Perhaps what is most amazing is the influance the NSA, GCHQ, et al had on manufactures in the first thirty years of the development of computers and in fact just how much better the UK government was at suppressing it than the US (with the result that few people knew what the UK’s input to computing during that time was). It needs to be noted that the influance of the US Government on that of the UK was fairly obvious and revolved around the significant debt that arose from before and after “lend-lease” which was just refered to as “war debt”.

Clive Robinson • March 19, 2012 6:23 AM

@ Brandon,

Of course, whether and how much the NSA and Google work together is up for speculation, no matter what Google claims.

The “work together” might noe be known to Google managment…

We know the NSA have “put employes in place” in major US telco’s prior to 9/11 to ease wire taps etc. In some cases this was unknown to much of the telco’s senior managment. This sort of behaviour goes back befor the NSA existed with the “Cable Companies” prior to US involvment in WWII.

@ AC2,

Anyone know why the power substation would be outside the security fence

Specificaly no but you can make reasonable guesses.

First off if you were an adversary who wanted to put the place out of commission using low tech or low personnel count you would go for either the power or data communications well outside of the security perimeter, and well away from roving foot patrols etc.

The comms can be fairly easily hidden or sent via radio, not so the power. Thus the secure compound has it’s own three day UPS. Further I would suspect there are also plans in place to gradually power the place down to critical functions only to extend that period and arangments made to ship in fuel and ship out functionality and personnel to other compounds in other places.

The cost of that tank proof compound fence is going to be a bit more than a few bucks/foot of chain link so it would be desirable to keep the perimeter as short as possible. And from the article it is not clear who owns and operates the substation… Which brings up the question of does it need to be secure at all, if not then it most definatly does not belong in the secure compound.

If you think of the substations function it is just the terminator to a very vulnerable and insecure power transmission system. Due to the UPS it serves no security function other than as the voltage step down for more secure cables into the compound and the UPS. And I suppose it should be said the UPS is a security function as it will be designed to meet EmSec requirments as a back stop to other EmSec equipment all of which will consume power one way or another.

So no the substation and it’s personnel should not be in the security compound. But that’s not to say it should not be protected to a much higher level than normal for a substation of that size or not run by security cleared personnel. It is a rather awkward “weak link” in that whilst over head high voltage cables can be easily damaged they are relativly easy and quick to bring back on line (days to weeks) compared to the substation which whilst equaly as vulnerable could take ten or more times to repair than the overhead cables.

Clive Robinson • March 21, 2012 1:56 AM

OFF Topic:

Some may remember one of the reasons Stuxnet was a bit of a shock was the code was signed with (believed to have been stolen) private keys coresponding to official code signing certificates.

Well it appears that malware developers are now doing the same because using stolen private keys makes it so much easier to get around AV software, because of a defect in the way the AV software works.

Some people may remember I have a very longstanding and significant dislike of code signing as a security measure because it is weak to the point of being “faux security” and can be got around in a number of ways. Some of which are,

1, Work for the organisation up stream of it’s code signing process, this is belived to have happened a number of times.
2, Get the organisation to sign your changes/additions to their code, this happens with mobile phone code, hardware driver code etc.
3, Find the private key of the public key, this has happened to a number of hardware/software organisations that have used short or weak generated random numbers for the keys.
4, Steal the improperly protected private key from the organisation, what is believed to have happened with Stuxnet.
5, Generate a false certificate by compromising a CA, we’ve seen this a number of times with browser certificates so it’s a reasonable assumption that code signing certificates are just as vulnerable.

Well Swiss company Conpavi AG’s private key for their digital certificate (issued by Symantec) appears to have been used by the writers of the “Mediyes trojan malware” according to Kaspersky Lab a few days ago.

Which has started an investigation as to how the private key was compromised. Symantec who signed the coresponding public key of the now compromised private key have revoked the certificate concerned.

One of the reasons using compramised private keys to code sign is desirable for malware is because of heuristics in AV software. The manager of Symantec security response Liam O’Murchu said,

“We have seen more [malware] being signed, sometimes with stolen certificates,” because “It lends an air of legitimacy to the file.”

He further indicated that, as Symantec has evolved its malware protection method, it has included a risk-based scoring system based on several indicators as a speed up process to determine if code ariving on a PC is benign or malevolent (ie just like any “fast track” process common in “security theater”).

The problem is that any code which is “Digitally Signed” gets a very significant advantage in the Symantec risk-based scoring system. What is not clear is if being “signed” gives it a pass from all further checking or not (Symantec aren’t saying).

Liam O’Murchu has acknowledged that, compromised private keys and the coresponding digital certificates are available in online criminal black markets, such as “carder sites” you would expect to find stolen credit card numbers etc.

He further acknowledged that if malware writers are effectively figuring out how to get around Symantec’s AV software scoring system it will need to be recalibrated…

http://www.pcworld.com/businesscenter/article/252099/stolen_encryption_key_compromised_symantec_certificate.html

Clive Robinson • March 21, 2012 10:48 PM

@ Kashmarek,

On the new Samsung privacy invading TV

Bad as the idea of the camera and mic is, it’s the licence agreement that is even more scary, basicaly allowing them to install third party apps and the third party to do what they want with the data/video/audio collected…

Now as we know from the US school who thought it was OK to remotly activate the camera etc on laptops issued to children and thus capture them in their bedroom in various states of undress, this does not go down well as it starts getting to a court room…

Also I’m not sure but it might well be in contravention of EU Privacy laws.

And in the UK if they capture an image no matter how fleetingly of a child (ie under 18) in a state of undress then that automaticaly activates the laws relating to “K1dd4 p0rn” irrespective of what the intended purpose of the image. The UK law is written to be so encompassing that you get the “law of unintended consequences” comming into play as proffessional photographas hired by the parents of children to take portrait style photos (like the traditional baby on fur rug) have been prosecuted. There was also a case of a girl taking a “naughty picture” of herself with her phone and sending it to her boyfriend she was threatend with prosecution untill “cooler heads” decided it “was not in the public interest to prosecute”.