The curtain of state secrecy which surrounds the type of government agency known primarily by initialisms is all-encompassing and long-lived, meaning that tech that is otherwise in the public domain remains top secret for many decades. Thus it’s fascinating when from time to time the skirts are lifted to reveal a glimpse of ankle, as has evidently been the case for a BBC piece dealing with the encrypted phones produced by GCHQ and used by Margaret Thatcher in the early 1980s. Sadly, it’s long on human interest and short on in-depth technology, but nevertheless from it can be deduced enough to work out how it most likely worked.
We’re told that it worked over a standard phone line and transmitted at 2.4 kilobytes per second, a digital data stream encoded using a paper tape key that was changed daily. If we were presented with this design spec to implement in a briefcase using 1980s components, we’d probably make an ADPCM (Adaptive Differential Pulse Code Modulation) system with an XOR encryption against the key, something we think would be well within the capabilities of early 1980s digital logic and microprocessors. We’re wondering whether the BBC have made a typo and that should be kilobits rather than kilobytes to work on a standard phone line.
No doubt there are people in the comments who could tell us if they were willing to break the Official Secrets Act, but we’d suggest they don’t risk their liberty by doing so. It’s worth noting though, that GCHQ have been known to show off some of their past glories, as in this 2019 exhibition at London’s Science Museum.
Schematics?
“Official Secrets Act” probably says no to that. But from the guess work in the article it is not that hard to guess the design blocks of it.
Why would you suggest an XOR test against the key? My naive idea would be to use AND logic, but I’m not familiar with this sort of digital encryption.
The truth table for XOR
0 XOR 0 is 0
1 XOR 0 is 1
0 XOR 1 is 1
1 XOR 1 is 0
At transmission site
01000001 Unencrypted message
01101011 100% random key
00101010 Encrypted Message ( Unencrypted Message XOR Key )
At reception site
00101010 Encrypted Message
01101011 100% random key (duplicate of key used at transmission site)
01000001 Decrypted message (Encrypted Message XOR Key)
Anyone can intercept the Encrypted Message, it does not matter because without knowing the exact random key used the encrypted message can never be decrypted, provided that the random key is truly random and there are only two copies of the key that is destroyed after one use only.
Ah, that makes sense. I’m familiar with the method of alphabetic key ciphers (like Vigenere’s), but I didn’t know how binary sequences were encoded. I see now that the XOR truth table allows both encryption and decryption without loss of information. Thanks!
Mrs. Thatcher was a medical marvel, how she survived with her blood pumped around her body by only by a stone is a mystery to this day.
Which reminds me of a joke what do Margaret Thatcher and Jimmy Savile have in common…?
For Jimmy Savile did it involve miners, and for Margaret Thatcher was is minors. Maybe I have that the wrong way way around, but they were both void of humanity.
You’re correct, I suppose the pun works better when spoken :-D
In 1970, Popular Electronics had a phone scrambler project. It used a Radio Shack speaker phone for the interface and housing. (You put the handset in the box, for acoustic coupling, so no connection to the phone line).
Simple audio inversion, a diode ring modulator fed by an oscillator (or as they suggested a radio for a more complex scrambling).
Easy to decode, if you knew what was going on, but most people wouldn’t have the knowledge. No word on what secrets needed a scrambler.
I’ve read that audio inversion was used for scrambling of some police radios when scrambling first came in.
The “best” scramblers have been in fiction books, a switch at one end turns a scrambler on, nothing needed at the other end.
AFAIK, During WWII, President Roosevelt communicated to Stalin through an audio inverter.
Experienced operators of the equipment learned re-invert the audio in their heads.
I hadn’t known itwent back that early.
It works fine, until the process becomes known, and some guy is selling kits in the back of magazines. Phones provide a level of security, since they have to be tapped, but radiois open to all listeners.
I remember that one, went something like “Listen to Ronnie and Maggies late night love chats” was by Loompanics or someone :-D
I believe the U.S. used something called Sigsaly, here’s the link.
https://en.wikipedia.org/wiki/SIGSALY
They did, but I’m pretty sure it was only ever installed in a few spots in Western Allied countries. Turing came up with some simpler systems when he was working with Claude Shannon at Bell Labs, I can’t remember whether any of those were actively used but it’s possible they might have trusted Uncle Joe with something which wasn’t the family jewels.
How you still install inverters on marine VHF radios, so the fisherman keep their spot secret. The other fisherman always tried to pay me to unscramble the radios for them. I wouldn’t do that but I would happily install an ADF.
“The curtain of state secrecy which surrounds the type of government agency known primarily by initialisms is all-encompassing and long-lived, meaning that tech that is otherwise in the public domain remains top secret for many decades. ”
“Q”‘s handiwork no doubt.
It was that sentence that tipped me off to it being the written work of [Jenny List].
B^)
I think this is to easy for Q/Q-branch.
At least he add a video link to this project.
This was a replay to Ostracus post
A standard 4-bit ADPCM can’t achieve 2400 bps, because that would result in 600 samples per second, with a top frequency of 300Hz, and the fundamental frequency of a woman’s voice. In practice, ADPCM is used with sample rates of about 8KHz giving a maximum audio frequency of 4KHz which captures most speech at a rate of 32k bps.
The US DOD released information about a 2400 bps voice codec, FIPS 137 in 1984, just a little later than this system is set, so it’s possible they used it internally in the UK and the US beforehand.
https://en.wikipedia.org/wiki/FIPS_137
CVSD is a really interesting modulator, which operated at typically 24k bps, 10x the quoted value here and much higher than we’d expect from audio in the 1980s.
https://en.wikipedia.org/wiki/Continuously_variable_slope_delta_modulation
Early 80s? Ensigma was demoing DSP-based speech encoding by about ’85 which used some sort of LPC: while the details (including bit rate) elude me it would not surprise me if GCHQ and their peers had something workable significantly earlier.