Funny looking line noise

[maradydd]

So this Ukrainian dude I follow on Twitter posted a link to this diplomatic cable from Kyrgyzstan, which has all kinds of goofy looking line noise in it. In fact, even the file name is jacked up -- it should be BISHKEK, not BISHIEK, which makes me think of some town in the frozen steppes whose sole export is yaoi manga

Comments

[etcet]

This idea is probably not clever, nor is it likely correct:

If those items were transmitted electronically, as opposed to via old-school analog (ie: telegraph), I guess it would be expected that any interference would almost have to be binary, wouldn't it? Why it's almost always discrete powers of two... and when it isn't, it's almost always a power of 2 plus 4.

hmmm...

Maybe they indicate the magnitude of the flicker in transmission strength, or duration of interference (though that latter doesn't really account for the negative values elegantly)?

[maradydd]

How are diplomatic cables actually sent these days? Telegrams certainly still do exist -- I received one last year, I think -- but I haven't the faintest idea how they're actually transmitted from station to station. I would assume it would have to be digitally, but perhaps I'm wrong

[jered]

Definitely not an OCR error. This pattern is definitely indicative of a non-ECCed digital channel; I don't think anything more suspicious is going on.

I saw something similar once when I had a SCSI hard drive with a bent pin. Wow, that sure made a mess. My first indication is when it identified itself as a "SEAGAUE". Yes, parallel SCSI did no checksumming...

[maradydd]

parallel SCSI did no checksumming...

*facepalm*

Yeah, I didn't think it was anything sketchy, though I'm still curious how far one could get toward pinpointing the actual error. Of course, without knowing the sequence of events in data transmission, that's nearly impossible to do, but it's nice to know that patterns like this signify some kind of error-correction fail.

In any case it highlights why it's important to use checksums when transmitting important files! I wonder if Wikileaks was even doing that.

[lightning_rose]

"I could tell you, but then I'd have to kill you."

Seriously, it looks too ham fisted to be deliberate steganography. Besides, steganography works best when no one suspects it's being used, and especially when it's invisible to casual inspection, such as in an image where it's easy to hide single bit errors.

[maradydd]

All true. I wasn't thinking steganography so much in the sense of concealing some message, more in the sense of origin detection, i.e., make N subtle changes in the wording or spacing of data stored in N different places, so that you can tell which place the data was leaked from, but this would be a silly and obvious way of going about that.

[mouser]

I'm going to say this with 25% confidence:

Something along the way is using some very old hardware or has a hardware design that doesn't have proper CRC checking in place. This hardware has an intermittent failure with memory, or a bus error.

I had a similar problem with a very old computer sorting records. Apparently when the unit heated up (well past the PowerOnSelfTest) one of the memory sticks would stop accepting an "ON" bit write operation. Screwed up enough data before it was found that I had to redo several months of work.

The other 75% possibility is that they are using a modem or other acoustic device with (again) CRC checking turned off, most likely because they've got the speed turned up beyond what is usable.

[anaisdjuna]

Ok. I'm going to pipe up and talk directly out of an orifice that has no exposure to direct sunlight, if you know what I mean. I like this post and I like spy-ish things and this thought came to me on the social plane... I have no geek powers beyond the clicking of mice.

Didn't this come from Kyrgyzstan? How old & funky is equipment over there? The Russian spies that got busted this summer were using old school, old stuff. Does anyone expect the clumsy Kryrgyzstan, steganographic inquisition?

Hee hee. I feel all tingly now.

[krfsm]

'Tis from the US embassy in Kyrg., not field agents...

[nibor]

Yea, I'm going to have to agree that it's almost certainly just a noisy line or some such. If the line is this noisy they might turn off checksumming just to get data through at a reasonable rate; it's close enough that the humans can figure it out, which is better than sending nothing. If this is being transmitted in ASCII (which is the only reason such a pattern would be meaningful) then single bit errors are going to look a lot like that. The existence of a "-1" offset makes me happy enough that the ones bit was even affected, and there appear to be a small number of dual-bit errors as well. I suppose you could run the statistics on the number of one bit errors and 2 bit errors to see if the error rates are reasonable (although it's possible that the chance of bit error on two nearby bits is dependent rather than independent) just to sanity check the theory.