Schneier on Security

Clive Robinson • August 18, 2012 4:01 AM

@ Chris W,

Gauss contains encrypted payload which can only be decrypted on the target machine

Encrypted payloads are not new but the technique as described for Gauss is quite fascinating and I had some thoughts about similar but not as sophisticated methhods many many years ago.

Way back the use of very very simple encryption (think XOR) was to hide the malware signiture from AV software. The idea was to produce a very very small piece of polymorphic decryption code as “the loader” and use a randomly generated key to encrypt the main functional part or “payload” of the malware.

The problem with this system was the decryption key had to be sent in “plain text” for the decryption loader to find it and in general it was very very short.

What I was looking to do was to improve the length of the “key” by using strings of OS code that were in “known positions” on the target PC so in effect I was sending not a key but a pointer to a key, but the pointer had to be obsficated as well (which is a subject for another day).

Now one problem I came up against was my loader was OS specific because MS used to change bits from version to version of DOS so the loader needed a checking mechanism to indicate if the decrypted code was good or bad so I had to embed some “known plaintext” into the payload.

Now for various reasons not just to keep the loader small I originaly used A5A5 (see BIOS/POST and loading I/O card software for why I chose those values). Although I doubt Gauss uses the same known plaintext it will almost certainly use some known plaintext and very probably at a fixed offset point.

Now Gauss supposadly uses ARC4 which is a stream cipher. Stream ciphers happen to be quite vulnerable to “known plaintext attacks” at fixed offsets as thay enable you to strip out the “known plaintext” and recover. part of the key stream which carries forward into other attacks on the stream generator.

Now the $64,000 question is,

“Have they obsficated the known plaintext?”

If not then that may well be the most promising opening.

If however it was me I’d obsficate it and the way to do it would be similar to the way Gauss generates the key.

That is if you work on the assumption it is looking for a particular software package on the target PC, you could use a chunk of that programs code segment as the “known plaintext”. Because you would only need to calculate the offset to the start point of the chosen string of data in the code segment. If the program was not in memory or of a different revision then with a very very high probability the “known plaintext” would be wrong.

But also let’s assume a false positive in the detection code causes a key to be generated which is incorrect and thus produces a garbage decode of the payload. The probability of a matching “known plaintext” would be vanisingly small. Thus it would be very very unlikley that a garbage decode would be treated as valid code causing the loader to try and execute it and causing the PC to crash.

Something else that is perhaps slightly odd about Gauss is it loads a narrow font onto the PC…

So the question is “why?”

I can think of several reasons some of which are,

1, It contains the known plaintext.
2, It contains data to obsficate or as KeyMat.
3, It has TEMPEST advantages.

It is known that some fonts have a very high spectral content and this is realy quite usefull when doing “van EcK” monitoring of VDU’s and other displays. The folks over at Cambridge Labs did some work on this ( http://www.cl.cam.ac.uk/~mgk25/ih98-tempest.pdf ) and actually demonstrated van Eck phreaking at a trade show some years back.

What ever the reason is I suspect that Guass is going to be quite a fruitful beast to disect and provide the up and coming security bods with some new things to think about (whilst those with “impressive beards” nod and say “Back in my day…” 😉

Clive Robinson • August 18, 2012 1:29 PM

OFF Topic:

And another bit of news that might be of interest,

Firstly though a bit of background on UK Gov Ministerial idiocy that has kept the more important news out of the main media outlets,

You may have heard that the UK Gov is kind of “throwing the toys out the pram” because Ecuador has decided to recognise that the WikiLeaks founder might just have a point about what will happen if he is extradited to Sweden and to then to the US where he may now be “held on suspicion without trial indefinatly”. So as he’s holed up in the Euadorian Consulate in London he is apparently making the UK Gov look bad and endangering the US-UK “special relationship”. So the UK Gov’s “William Hague” has rather foolishly said the UK Gov will remove the “diplomatic status” from the Ecuadorian consulate and send the police in to search for and arrest the wikileakes founder all of which is against the Vienna Treaty… All in all it would set a bad president in that if done once by the UK it leaves all UK diplomatic buildings vulnerable around the world. So well done Mr Hague you’ve taken a cheap shot at your own foot and potentialy endangered the life of every person who has UK Diplomatic status. It’s interesting to note that a number of ex-UK senior diplomats have come out and said in best diplomatic terms “he’s an idiot”.

So baack to the real technology news behind that… whilst all this stuff and nonsense was happening Wikileaks. dropped some new papers into the public domain about something called “Trapwire”. It would appear to be a plan to link all CCTV cameras upto a central DB via face recognition software to track the movment of all individuals either activly or in retrospect.

There are some “hysterical reports” out there about how ‘life is now a goldfish bowl” and others that say the technology is not currently possible and it’s just a ruse to slurp off more tax payer dollars to the chosen few.

Either way this artical gives some background on Trapwire and some other things we should be keeping a watching brief on,

http://m.techcrunch.com/2012/08/18/move-along-no-panopticon-to-see-here/