unforum

[bwochinski]

[SOLVED] Contact1976.wav

[Rogi Ocnorb]

While it would be easy to create a file like this, getting the information out is another matter, entirely. We have no clock with which to define the pulse widths of each distinct "bit". It may be possible to derive the clock frequency using audio software, but I don't know how to do that.
Can anyone think of a way to capture the distinct bits in the signal when they repeat?

Example:

[Ehsan]

Just a brief to explain what we've been doing with this one for the past 2 days.. I've been discussing this with xnbomb and he did most of the coding on this stuff..

Basically we know this is TTY. It's obvious from the 1400/1800 frequencies which is the US standard.

Unfortunately, all attempts by everyone working on this to make it work with any programs we can download failed. Everyone just keeps getting garbage.

So we just looking at the file itself.. this is the spectral view

Now manually looking at that and converting to binary we get:

100000001110110010001000100100001000100011101000100000001010110010010000101010001000000010101000

Then, broken down into groups of 8 bits that's:

10000000
11101100
10001000
10010000
10001000
11101000
10000000
10101100
10010000
10101000
10000000
10101000

But it's not standard ASCII.. notice that each group starts with 1 and ends with 00. That's again a TTY standard, 1 start bit and 2 stop bits around a 5 bit message. So we just take the middle 5 bits from each line:

00000 11011 00010 00100 00010 11010 00000 01011 00100 01010 00000 01010

Now what SHOULD have happened is that because this is TTY, we use the Baudot table (Also known as ITA2 alphabet) to convert the above into letters..

Of course things are never that easy.. It doesn't work. We get garbage output. At least it doesn't have any recognizable words or even letters.

Now this could be interpreted in a different way, the 0's and 1's could be inverted. Again this doesn't give any interesting results.

Also it could be flipped right to left. This in combination with inverting doesn't help.

Of course it's hard to extract the entire binary from the wav by just looking at it, and we have no idea how to program something like this, so xnbomb used the output from one of the programs that worked, took that letters and tried to figure out which baudot table they are using. After that he reverse engineered it back into 5 bit binary. I wrote a program that takes that as input, tries 3 different variations of baudot tables found on the internet, in addition to reversing the direction and inverting the bits. 12 different output, but none of them are interesting enough or make any sense to warrant further investigation.. we still end up with garbage text.

I don't think we should look at that reverse engineered binary though because we might've missed something. What's 100% is the 5 bit binary posted above:

00000 11011 00010 00100 00010 11010 00000 01011 00100 01010 00000 01010

This is how the file starts.. now knowing all of the above, we need to figure out what these are. They SHOULD be Baudot, but as shown that's not useful. Maybe some sort of double encoding. We even thought substitution cipher but we're not even getting proper words.. control characters, line breaks, and spaces are all over the place.. so I don't know what else to do with this...

Still working on it though...

[rose]

In case this is useful I am reposting the three pages of the report here. I don't know what type of file they were originally.


page one is attached and was found in Milwaukee

page two is here as a tif file and was found in Indiana

page 3 is here as a jpg file and was found in Indiana

[Lucy]

Newbie post...

Considering Ehsan's educated comments above and the general messiness of my results, I'm guessing the information I'm about to add is probably useless. Anyway, I ran the wav through some tty software (from http://www.tversoft.com/company/DXsoft/Download-CallTTY.html) and came out with the following:

[tom_in_bristol]

It seems like we can extract binary data from the audio, and it fits the baudot/5-bit character format. So this is key.

It might be worth considering the fact that any message hidden within may not start at the start of the wav file - it might be wrapped up in the middle of the file, or possibly start at the end and be encoded in reverse (i believe traditional tickertape tty messages sometimes needed each character byte to be reversed)

In an ideal world we would have a program to extract the raw binary (including start and stop bits) and then we would be more free to play with it... I can't find one though. Unfortunately reverse engineering baudot might be unreliable since there is no guarantee it is a true transcription of the raw data including possible long strings of idle bits etc etc.

Ho hum.

[tom_in_bristol]

[bwochinski]

Well, I'm only a minute in, but I'll post what I've got so you guys can start working with it.

The linebreaks are pretty arbitrary, about every 1.5 seconds, as that's all that fits on my screen at once, so don't worry about those.

Anyone feel like double checking my work??

[rose]

Bwoch -do you mind explaining how you got the translation? And thanks by the way.

[bwochinski]

Well Rose, if you look at my first post in this thread and look at the spectrum view, you can see the bold pink horizontal lines. The higher lines are 1s and the lower lines are 0s. Also in that image you can see vertical bands.

So what I do is I open the file in audacity and view it in the spectrum mode. I count each of the light blue vertical bands and type that many ones or zeros. Then I scroll to the right (about 1.5 seconds each "scroll") and repeat. It's actually not that bad... pretty mindless.... really....

*bwochinski stares into space

[xnbomb]

bipbipbipBIPbipBIPBIPBIPbipbipBIPBIPbipbipBIPBIP

[rose]

[tom_in_bristol]

Just to save everyone doing it, i have:

Taken the sequence of 5-bit characters
Converted the binary to decimal (to make the whole thing easier to work with)
The result is attached.

Below is a quick frequency count of the decimal values, 0 to 31.
It seems there are two large values, two nil values and the rest are neither here nor there. Make of that what you will:

00 0
01 63
02 1
03 74
04 45
05 46
06 68
07 63
08 0
09 53
10 62
11 51
12 66
13 63
14 58
15 45
16 61
17 23
18 29
19 55
20 37
21 59
22 45
23 49
24 39
25 55
26 67
27 331
28 56
29 46
30 44
31 339

[Rogi Ocnorb]

In the interests of getting others involved in this, I'm attaching a rich text format document that shows the entire 5bit stream that xnbomb has provided (You rock!). I've applied the decryption for five "Baudot" code tables I have found on the internet:
Murray
Teletype(TTY)
TDD
U.S.
CCITT-2

Letters don't vary from one to the next so they are only included once.

You can walk through the data from top to bottom and see how the decoding is done. The yellow highlighted rows are the control "shifts" xnbomb referenced that occur every 72 bytes.

Windows folks can use wordpad to open the file.
If anyone finds other 5bit code tables, let me know where.

EDIT: Doc was too big, so I'm zipping it.

EDIT: Found another code list called Illiac.
But it's giving really different results (Lots of spaces, though). Posting in case anyone else wants to play with it.

[xnbomb]

Now again, only better ...