Schneier on Security

Clive Robinson • May 2, 2012 11:28 PM

@ DoubleDown,

5) This whole story is a fabrication put out by some intelligence agency as some sort of misdirection campaign. (This type of thing happens all the time — more than most people suspect. The media is a powerful weapon for the TLA’s).

Hmm you did not go the extra step to option six…

Instead of “intelligence agency” and “TLA” use “Terrorist organisation”.

As I’ve noted before “Cpl. Hotfoot” “Cpt. Thunderpants” and the “toner cartridge duo” are terrorist attacks designed to fail as physical attacks but succeed as publicity attacks and a reminder that the “bogie man” is still out there.

They have the advantage of keeping the taps flowing on funding, whilst also waving a flag to attract people to the cause, as well as making the US and other countries security forces run around and their respective politicos destroying their economies and represing their people. A nice Win-Win for the Terrorists Organisation, the security forces and their political masters…

Ask yourself what has the terrorist organisation lost?

Absolutly nothing apart from a supposed courier who could well been a patsy set up to fall. From what has been reported (and that’s next to nil) the documents were either low level intel you could put together from the Internet or “hey how about this” fantasy planning ideas along the lines of a security theater contest…

With regards your fourth option, you almost got it 😉 but with any “disinformation operation” there does not need to be another operation to cover up.

For who ever set this up (and I’m thinking it is definatly some kind of setup) this has done realy quite nicely on it’s own just in time for the aniversary of OBL’s death. The “porn” angle ties in nicely with what was supposadly found in the OBL’s compound in Pakistan and the stego aspect is one of those long running “Oh My God” things that has also been around for a long time.

The story is “to pat” and that is a big big question mark. OBL has been dead for a year, and this style of terrorists communications method was clerly highlighted at the time. So ask yourself the question why have they not changed it as it would obviously be a “red flag” for even the TSA?

Oh and why were the documents so quickly recovered?

Even with the level of distrust they (supposadly) may have for “things western” we have good reason to believe that OBL’s organisation had strong links to people within the Pakistani Intelligence Service (ISI) and thus secure comms to a much higher level would have been well known to the ISI personnel and thus to OBL & Co…

@ Bruce,

The reports are still early, but it seems that a bunch of terrorist planning documents were found bunch of terrorist planning documents were found embedded in a digital file of a porn movie

I’ve had a little dig around and although there are lots of “Oh My God” reports, they all appear to go back to the one investigative journalists article in the German weekly newspaper.

Now the original German article actually does not appear to say a lot (as far as my poor understanding of German alows 😉 but it appears certainly on the technical asspects to be a good deal less than has been subsiquently reported.

The story appears to be developing a “life of it’s own” like any good “Chinese whisper” and is certainly making the journolists name well known around the world (which won’t hurt sales of his book). However I would have a look into his track record on middle east terrorist organisation stories to see what the independant and verifiable facts behind them are.

So I’d remain “firmly sceptical” unless a good deal more corroborating information from other independent sources comes to light.

Clive Robinson • May 3, 2012 2:04 AM

@ Osborn Greene, O W,

Encryption before steg makes it harder to do good steg, and easier to detect. Stuff that’s too random sticks out like a sore thumb

Take a whole rgb snapshot then compare the light(hue) fluctuation, as encryption would have large difference say 0x31 ->0xff

These problems are trivialy solved with a tailored expansion / inflation algorithm that adds “tailored redundancy”.

A very simple example would be take your encrypted output and break it into small blocks of 4-6 bits. To each value in the small block set assign one or more blocks of 8-16 bits.

You asign the bit range and number of large bit blocks to your small bit blocks to match the desired distrubution you require. You likewise select the individual large blocks to reflect other “trends” that change with time etc.

This has been known about and used since atleast WWII where the output of a “number one time pad” was then “remapped” to give a frequency distribution similar to that of other ciphers in use, so that the enemy could not easily tell OTP traffic from the weak “poem cipher” traffic or “dummy traffic” and thus could not use it for traffic analysis and it would also have the consiquence of tying up enemy cryptographers which were a very scarce resource.

Clive Robinson • May 3, 2012 2:37 PM

@ Alex,

The story stinks… …and guess what, the journalist who writes about the story is known to pimp up his stories with “insider knowledge” on “the evil terrorists”.

I would agree with your first point, and infact I’m not even sure there actualy is a story at all, as there appears to be little or no supporting evidence from other journalists and as I noted above the story appears to be based on next to nothing that is verifiable or rumours that have done the rounds a number of times including a year ago when OBL was killed.

However your second point “pimping” may be a little strong, it’s difficult to tell.

Supposadly he is an expert in the field, but then most of the so called experts the US Gov etc have used in court etc turn out to be “web site readers” not actual investigators testing and evaluating the accuracy of each item against multiple independent sources etc.

As I noted above my German is far from what it could be, but I have my doubts which is why I advised Bruce to look into the journalists other articles and book.

And I’ll repeate what I concluded,

So I’d remain “firmly sceptical” unless a good deal more corroborating information from other independent sources comes to light.

That is what he may be saying is true, but I’ve no way currently of verifying it, and I would be loath to draw any conclusions as both sides of this “great game” are resorting to “smoke and mirrors” and “faux action”. And as I noted above this story has many of the hallmarks of a “disinformation operation” that either side would benifit from greatly.

Clive Robinson • May 3, 2012 10:05 PM

@ O W,

I didn’t explain it right, its not frequency distribution. You get a overall light value(for sun,cloudy day) and then you check one pixel to the next and see if theres a large jump in light If they used alphanumeric encryption, which the text is in, and just use say uppercase the light values wont jump from dark to light. Most if no all encryption has uneven mapping

What you are describing is a numeric range and rate of change which my previously described method solves as well.

Overly simplisticaly the “small bit size blocks” set is small and only has 16 members for a 4bit block size from the original ciphertext. These 16 members map into 256 or more members of the “large bit size blocks” which actually get embedded in the carrier file. Thus you get the redundancy to select more than one value for each small bit size block from the ciphertext.

The first thing you do is analyses the carrier file to get the required “range” distribution you then make the map to translate the small blocks to the large blocks to match this distribution.

Overly simplisticaly each of the 16 members of the cipher text set would have a minimum of 4 of the 256 carrier file blocks asigned to it. Importantly these would be spread across the range values not given adjacent values. Thus the actual value of the carrier file block used to represent the ciphertext block would be selected by examining the carrier file pixel values adjacent to the one selected to hide the ciphertext block. There are a number of ways you could do this but again overly simplisticaly you could use a running low pass filter with the window covering the previous six pixels and next pixel with appropriate weighting. The value selected from those that map to the cihpertext value would be the one closest to the output from the low pass filter.

You could look on the process in the same way you reduce ISI (inter symbol interferance) in a multi-level communications system.

However as I’ve noted this is an overly simplistic way to do it, so that the basic idea can be seen. In reality because the carrier file charecteristics change on a continuous basis you would use a dynamic mapping system that had a window size appropriate to the carrier files ongoing charecteristics (think a single video frame and a two dimensional distance related estimator for the pixel).

Personaly I’d not use a mapping process at all and simply use a synthetic noise source modulated on a bit by bit basis that gets added across the entire window size in a simillar way to a Low Probability of Detection (LPD) radio system works, have a look at Spread Spectrum Communications systems and their “coding gain” which was used as the basis for many “Digital Watermarking” systems.

I could go on at further length to describe a more concrete example if you wish but it would take up quite a bit of blog space which I’m loathed to do.