[FRIAM] privacy games

[uǝlƃ ☣]

On 5/26/20 8:22 AM, Steve Smith wrote:
> 1st order)   When we encounter a signal (use text stream as a familiar example) we may or may not recognize that there is obfuscated meaning in that stream.  In the common example, of course, the stream usually looks like pure gibberish...  having an *apparent* high entropy.   Attempts to decode the stream usually involve seeking transforms which yield a low entropy or high information content.   Ideally, yielding a very specific, highly unambiguous text stream which is not only recognizeable to the decoder but possibly directly meaningful.   In the classic imagined examples, we have spies and counter spies attempting to pass messages and intercept/decode those messages, etc.    This is where the specific technical term /Steganography/ takes on interest and I think alludes to or defines your 3rd order?  I'm not trying to impute specific meaning that you didn't intend, just looking to tease out the language you are seeking to use and align it with existing lexicons which may
> or may not be fully apt for what you are getting at.

I don't quite understand how you're criticizing the 1st order idea. Is it only to say that I've mixed up my orders? ... that I'm confusing 1st with 3rd?  If so, then yes, I probably am. The particular examples of steganography I identified were "hiding" a QR code in an image (or vice versa) and hiding 2 images inside 1. And I used these as a foil to talk about the combinatorial explosion. So, while steganography, in general, and even these 2 examples are *not* purely 1st order, they help (I think) highlight the 1-many mapping. The more strictly 1st order demonstration of the 1-many mapping was the string comprehension example. Sorry for the confusion.

> 2nd order)  I am literally not clear on what the implications of many-to-many are here.   1st order... one-to-many would seem to imply that the *decoder* is searching through the space of possible decodings (combinatoric) for the presumed singular encoder, but it also implies that the *encoder*  is choosing from a similarly large number of *encodings*.   Perhaps you are alluding to the case where some encodings can be decoded by more than one decoder or in some cases, multiple encoders can be decoded by the same decoder?  I'm not sure what you are getting at, though I *am* confident that you are getting at somethings specific that I'm simply missing (so far).

The most important point to 2nd order privacy is the ability to use composite *encoders* whose [quasi]independence/[quasi]orthogonality is conserved across the many-many mapping. If there's a pattern on the surface that has been generated by a composite encoder-plex (with invariant orthogonality across the map), then you can use *either* decoder1 or decoder2 and get an independent decoding, that stands on its own. This is akin to your idea that your 20-year-old decoding of Pirsig was/is still just as proper as your recent decoding.

And 2nd order says it's *irrelevant* whether Pirsig did this intentionally or not. What's relevant is that the map is many-many ... i.e. allows for composite encoders with this invariant property across the encode-decode map.

> 3rd order and beyond)  I don't know the technical implications in cryptography for iterated encodings by different means.  My own preferred examples have multiple encodings being very different in quality... and in particular semantic and socio-cultural encodings of a message as implied by your reference to Moorcock/Joyce and poetry in general.

Excellent! Bring up "quality" is probably important, not least because that's the heart and soul of the hard problem ... the ultimate metaphysical assertion of privacy. But I do think this is beyond 3rd order. Maybe it's the 4th and final order. It introduces an additional boundary. And (hearkening back to your and Jon's suggestion that the maps have to be lossless and maybe invertible) we'll have to start talking about whether there is a closure of the spaces beyond the boundaries. But I feel like you've jumped ahead and your inclusion of "and beyond" is making us sloppy.

> FWIW, I would like to suggest that not all obfuscation is adversarial in the strong sense. 

I agree completely. I tried to say that by mentioning positively intentioned meta-games. Another of my favorite novels is "The Magus" by John Fowles. But I think pretty much any mentor-mentee meta-game falls in the same category.

-- 
☣ uǝlƃ