Rendered at 18:02:38 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
Merrill 24 hours ago [-]
A 1983 review of this remarkable system for the Institute of Electrical and Electronic Engineers (IEEE) attributes no fewer than eight "firsts" to SIGSALY. They are as follows:
1) The first realization of enciphered telephony
2) The first quantized speech transmission
3) The first transmission of speech by Pulse Code Modulation (PCM)
4) The first use of companded PCM
5) The first examples of multilevel Frequency Shift Keying (FSK)
6) The first useful realization of speech bandwidth compression
7) The first use of FSK - FDM (Frequency Shift Keying-Frequency Division Multiplex) as a viable transmission method over a fading medium
8) The first use of a multilevel "eye pattern" to adjust the sampling intervals (a new, and important, instrumentation technique)
OP here, yes, good write up, although I wish someone would code up a simulation that showed it and allowed us to hear what this telephone sounded like to FDR and Churchill.
Oh, wow, that's a great page with a lot more technical detail than what I originally posted ^^ highly recommended.
teddyh 21 hours ago [-]
> I think that the relevant patents remained classified until 1975.
If they were classified, they can’t have worked as patents. Patents, by their nature, must be open and available for others to read.
solyin 21 hours ago [-]
No they don't. There's an exemption in US law for the NSA/the government to file hidden patents which only become revealed when someone else files the same/a similar patent.
From what I heard when I was a patent examiner, there was a secure area of the patent office where these classified patents would be examined. I never knew where it was but there were rumors.
Animats 21 hours ago [-]
Here's the technical description, from NSA.[1]
It was really an awful kludge. Today, you'd think, digitize audio into a bit stream, encrypt bitstream. That's not how it worked. It was somewhat similar to the older A-3 analog scrambler, which split up audio into a number of bands and rearranged them, changing the rearrangement occasionally. In SIGSALY, incoming audio was filtered out into 10 frequency bands.
There was also a dominant pitch detector and a "hiss" detector for unvoiced content.
That created 12 analog signals to encrypt. Those signals were much lower bandwidth than the original speech. The patent seems to indicate that the frequency band signals were merely encoded as above or below a threshold. Not a full A/D, just a 1-bit threshold. As late as 1973, people were still trying vocoder-like technologies, but it became easier to work on audio in the digital domain and that died out.[2]
The keying system changed the key every 20ms. It's not clear from the description whether the digitization of the 12 signals was also one sample for 20ms, but it probably was. If you used the same key bits for more than one analog sample, that would create a vulnerability. It's a one time key system, and those only work against serious opposition if the key is only used once. See VENONA.
The articles indicate that synchronizing the key stream, which was on phonograph records, was a huge synchronization headache. Keeping signals in sync was really hard, and was part of the curse of analog TV, with horizontal and vertical hold controls.
What this looks like is a system which had to work regardless of cost. So it was put together out of known building blocks. Straight digitization of audio was probably considered, especially since Alec Reeves, who had demoed pulse code modulation in 1937, was involved.
But it was a high-risk technology. They would have had to invent a good A/D converter and a decent modem, which would have added technical risk to the project.
It's one of those systems of historical interest because it was too soon. You couldn't really do the job in the straightforward way, but it needed to be done, resulting in a magnificent kludge.
I'm not sure whether or not I'd call it a kludge, but the workings are totally unlike the A-3 scrambler.
Splitting audio into bands and voiced/unvoiced content is basically how many "modern" low-bandwidth voice coding techniques like LPC, MELP, or CLEP work (see https://en.wikipedia.org/wiki/Linear_predictive_coding). The point of that processing is to compress, not encipher, the voice data.
Analog scramblers, like A-3, by contrast, shift the frequency content of the speech as the means of encipherment.
The point of the SIGSALLY voice coder is simply to keep the amount of key material for the one time pad reasonable. Even with that compression, each OTP record lasted only 12 minutes. MELP, which is a current DoD vocoder standard for use in secure voice channels, produces intelligible speech rates as low as 600 bps.
The part that is definitely a kluge is the key synchronization method. It seems that they did not use a PLL (which had already been invented by that point) to keep both ends in sync and instead relied on highly precise frequency references and turntables driven by synchronous motors with hundreds of poles.
NitpickLawyer 22 hours ago [-]
> a pair of vinyl phonograph records, each containing identical-but-random noise.
> These were the (single-use) keys to the encryption and decryption process.
> Then they set their identical records spinning at the exact same moment. On each end, the voices of these two world leaders are threaded through the SIGSALY terminal, mixed with a layer of noise, sent across the ocean via shortwave radio, then re-synthesized in a reverse process.
> Anyone attempting to listen in on the calls would just hear a buzzing sound, which is how the system earned one of its nicknames: “The Green Hornet,”
> At the end of each conversation, the records were destroyed to protect against potential future decryption.
I wonder if with todays tech we could decrypt this. This sounds really close to a one-time pad (if the conversation was kept shorter than each record) so .. probably not? Unless their random noise wasn't as random, or if the mixing of random + voice would leave other tell-tale signs?
vessenes 20 hours ago [-]
Definitely a one time pad - and therefore the answer to your question is "it depends on the randomness used". Sound sometime has surprising decryption properties in that starting and stopping long runs of data isn't always obscured properly by some schemes, so I'd also say "it depends on the exact scheme used for the frequency shifting."
This was at the dawn of the age of modern information theory, built on the shoulders of giants, but there are often some subtleties.
An example: if we imagine the encryption is perturbing a waveform +/- 30% of its frequency and amplitude at any one time, the encrypted bitstream is likely totally intelligible to the ear when replayed, even if reconstructing the exact bitstream is difficult.
Another example - if you make a block cypher that relies on prior blocks to seed randomness, then encrypting say 1,000 sequential blocks of zeros is very likely going to have some structure to it unless you've specifically designed it not to.
I'd love to hear some waveforms of what these calls sounded like encrypted -- it would be a fun project to work on decrypting them with modern tech.
fennec-posix 20 hours ago [-]
Might be able to re-produce the signal in GNU Radio, I think there might be enough information in the Wikipedia article about how the modulation and encryption worked to get something similar going. The actual voice that came out was barely understandable lol: https://www.youtube.com/watch?v=OvR4qK0B--w&t=517s
saghm 22 hours ago [-]
> Each terminal cost $1 million (at 1943 values) and weighed more than 50,000 kg (50 tonnes). The first one was housed in the Pentagon, with an extension to the White House. A further 11 terminals were installed at different sites around the world, including one in the sub-sub-basement of Selfridges department store in central London
I've learned that I tend to underestimate how large inflation is for historical values, so I looked this up. $1 million in 1943 is apparently just under $19.5 billion today, which means the full set cost $234 billion. I'm not sure if the US foot the bill for all of them or not, but to make it more concrete, 44 of these would have cost more than entire budget for the Department of Defense for this year.
vessenes 20 hours ago [-]
Others noted this should be $19m, not billion. That said, often a more useful analysis is percent of GDP -- this can give a sense of what fraction of effort of the total economy went into it. On those terms, 1943 GDP: $200bn, 2025 GDP: $30tn, for an increase of 150x.
So, depending on how you count, these were $19-150m each in today's dollars, or between 1/10 and 1/3 of an ASML EUV machine.
saghm 18 hours ago [-]
Yeah, I honestly wouldn't have commented at all if I had avoided such an embarrassing reading comprehension mistake, since the correct number is not particularly interesting.
My lesson for today is not to trust my ability to count comma separated digit groups.
z3dd 22 hours ago [-]
1 mil seems like an arbitrary or at least rounded number, and also it's not like you could get one on the market, so maybe it actually didn't cost a million
leonidasrup 22 hours ago [-]
For comparison, the unit cost of Boeing B-29 Superfortress (manufactured 1943–1946) was $639,188. In total 3,970 B-29s were build.
Crap, today is not my day for reading comprehension I guess (combined with overcorrection for my anticipated error). Too late for me to edit, but thank you for pointing this out.
ahmedfromtunis 23 hours ago [-]
I know this is a bit offtopic, but shouldn't the title say "the President of the United States," since 1) Churchill is UK's Prime Minister, and Roosevelt is not his President and 2) the publication on which this article appears is from the UK?
saghm 22 hours ago [-]
If the article itself (seemingly from the UK, as you point out) is inclined to assume that it's fine to describe it this way, I don't see why those of us from the rest of the world (which I'm inferring from your username is a group you're a part of) should object.
jhbadger 22 hours ago [-]
Even the original UK article just says "the President". Things may change as the US may not play the leading role in this century as it did the last, but despite many countries having presidents (although not the UK), it is still understood that "the President" means POTUS.
23 hours ago [-]
beejiu 21 hours ago [-]
The UK doesn't have a president, so it normally refers to the US president in general context.
dfxm12 22 hours ago [-]
I think historic context provides enough clarity for readers of the website of the imperial war museum.
I also think the headline as is would be good enough even for people who might not have learned about ww2 yet. A headline is not meant to be a complete summary after all...
This is briefly mentioned in Neal Stephenson's Cryptonomicon. The book is well worth a read if you don't mind viewpoint shifts and slightly hokey historical spec-fic.
nicbou 14 hours ago [-]
I reread it earlier this year. It's a very fun book.
ErroneousBosh 20 hours ago [-]
> slightly hokey historical spec-fic.
It's a slightly alternate universe. It's not quite our world, it just looks kind of like it from a distance. Qwghlm isn't really Skye and Lewis, it's like if you kind of wodged the Falklands and the Isle of Man into the same space.
Some of the people are kind of the same but not quite.
But from way over here, you'd be hard pushed to spot the differences, unless you looked annoyingly closely.
As mentioned, that had a tough sync problem. SIGSALY changed its key every 50ms. The mechanical synching of two turntables within that range was possible. That worked because the audio was being compressed pretty hard before encryption.
Delilah, which works by using a modular analog adder to add signal and key noise, would need a new analog key value for every audio sample. Probably 3KHz or so, or about 300usec. That's beyond the precision of sync which can be achieved reliably by mechanical means.
A little bit of analog buffer would help.
It wouldn't be a problem for transistor era electronics. You need something like an analog shift register (see "bucket-brigade device"[1]) Those were not developed until the 1970s, and, for a while, appeared in guitar pedals, reverbs, and other simple audio processing devices. If you had to do that with tubes, you'd have a whole rack of tube delay line.
Wild, I literally just this week started reading “How to wreck a nice beach” [0] which is already turning out to be a pretty entertaining book. And then this turns up. Thanks HN!
Haha I thought of the same book as soon as I saw this HN topic!
stmw 17 hours ago [-]
OP here - I'm so glad this got picked up and people are enjoying it. So many excellent additional details, like the list of firsts and other history.
AlexCoventry 24 hours ago [-]
IIRC from Prof: Alan Turing Decoded, Turing was sent to the US to work on this for a while, but wasn't able to contribute much, because the Americans didn't really trust him.
hydrogen7800 23 hours ago [-]
There was a 99% invisible episode on this [0]. I recall there being some good audio examples.
For anyone interested in this stuff - “Battle of Wits” is a great book about cryptanalysis in WW2.
tehnoslow 21 hours ago [-]
Thanks, now I know what I'm going to read next
daveevad 18 hours ago [-]
Anyone decoded SIGSALY correctly yet, is it signals ally?
gnabgib 18 hours ago [-]
Read on (in TFA):
> It came to be known as SIGSALY
daveevad 18 hours ago [-]
Didn't read the article (traditionally) but did read the Toll Trilogy on PACWWII. There are a bunch of these types of strings in Army and Navy history. It isn't obvious to me yet the general stenography, style of these things.
idontwantthis 22 hours ago [-]
Does anyone know how they synchronized the time? Was a radio signal fast enough?
phire 14 hours ago [-]
Based on this NSA documentation [1], they didn't actually synchronise the clocks with each other. They just tuned each machine's quartz oscillator with the nearest national time source. They were mostly dialling in the oscillator frequency here, not the exact time, so it didn't really matter they were synchronising to different national time standards.
Besides, even with the automated start system, the actual spin-up time of the turntables would have been a little chaotic. I suspect they were only aiming to get the four turntables within about 100ms of each other. The decrypted signal itself could then be used for the final fine-tuning of the timings (which also included adjusting for any speed-of-light delays)
If you weren't within 20ms, the operator would only hear noise (like a radio). Once they were within ~20ms, the operator would start hearing silence and would tune the phase until all noise went away.
You don't really need exact synchronisation because like a radio, the operator could just scrub the tuning to find the silence.
The synchronised clocks meant that once the operators had locked on, it would stay locked for a long time.
I don't, but by that point it was a solve problem - speed-of-light is certainly fast enough, especially with optional pre-bias correction term based on distance. 300,000 km/s, say 6000 km, approx 20ms.. It's also oddly similar to the 50 Hz sample rate.
vessenes 20 hours ago [-]
Wikipedia says 'precise clocks' which I would guess was quartz clocks on the back of radio synchronization. It was definitely well within the capabilities of the scientists of this period to precisely calibrate against radio wave time delays from London, Australia, etc.
1) The first realization of enciphered telephony
2) The first quantized speech transmission
3) The first transmission of speech by Pulse Code Modulation (PCM)
4) The first use of companded PCM
5) The first examples of multilevel Frequency Shift Keying (FSK)
6) The first useful realization of speech bandwidth compression
7) The first use of FSK - FDM (Frequency Shift Keying-Frequency Division Multiplex) as a viable transmission method over a fading medium
8) The first use of a multilevel "eye pattern" to adjust the sampling intervals (a new, and important, instrumentation technique)
From https://www.jproc.ca/crypto/sigsaly1.html
I think that the relevant patents remained classified until 1975.
(I see now that it was already linked elsewhere in the the thread.)
Pretty impressive for cross atlantic one-time pad voice comms in an analogue era, I think. It's amazing what money and smart people can accomplish.
Also found this fantastic page: https://www.cryptomuseum.com/crypto/usa/sigsaly/
If they were classified, they can’t have worked as patents. Patents, by their nature, must be open and available for others to read.
Invention Secrecy Act of 1952. https://en.wikipedia.org/wiki/Invention_Secrecy_Act
It was really an awful kludge. Today, you'd think, digitize audio into a bit stream, encrypt bitstream. That's not how it worked. It was somewhat similar to the older A-3 analog scrambler, which split up audio into a number of bands and rearranged them, changing the rearrangement occasionally. In SIGSALY, incoming audio was filtered out into 10 frequency bands. There was also a dominant pitch detector and a "hiss" detector for unvoiced content. That created 12 analog signals to encrypt. Those signals were much lower bandwidth than the original speech. The patent seems to indicate that the frequency band signals were merely encoded as above or below a threshold. Not a full A/D, just a 1-bit threshold. As late as 1973, people were still trying vocoder-like technologies, but it became easier to work on audio in the digital domain and that died out.[2]
The keying system changed the key every 20ms. It's not clear from the description whether the digitization of the 12 signals was also one sample for 20ms, but it probably was. If you used the same key bits for more than one analog sample, that would create a vulnerability. It's a one time key system, and those only work against serious opposition if the key is only used once. See VENONA.
The articles indicate that synchronizing the key stream, which was on phonograph records, was a huge synchronization headache. Keeping signals in sync was really hard, and was part of the curse of analog TV, with horizontal and vertical hold controls.
What this looks like is a system which had to work regardless of cost. So it was put together out of known building blocks. Straight digitization of audio was probably considered, especially since Alec Reeves, who had demoed pulse code modulation in 1937, was involved. But it was a high-risk technology. They would have had to invent a good A/D converter and a decent modem, which would have added technical risk to the project.
It's one of those systems of historical interest because it was too soon. You couldn't really do the job in the straightforward way, but it needed to be done, resulting in a magnificent kludge.
[1] https://www.nsa.gov/portals/75/documents/about/cryptologic-h...
[2] https://www.youtube.com/watch?v=UfpUexWJqb4
Splitting audio into bands and voiced/unvoiced content is basically how many "modern" low-bandwidth voice coding techniques like LPC, MELP, or CLEP work (see https://en.wikipedia.org/wiki/Linear_predictive_coding). The point of that processing is to compress, not encipher, the voice data.
Analog scramblers, like A-3, by contrast, shift the frequency content of the speech as the means of encipherment.
The point of the SIGSALLY voice coder is simply to keep the amount of key material for the one time pad reasonable. Even with that compression, each OTP record lasted only 12 minutes. MELP, which is a current DoD vocoder standard for use in secure voice channels, produces intelligible speech rates as low as 600 bps.
The part that is definitely a kluge is the key synchronization method. It seems that they did not use a PLL (which had already been invented by that point) to keep both ends in sync and instead relied on highly precise frequency references and turntables driven by synchronous motors with hundreds of poles.
> Then they set their identical records spinning at the exact same moment. On each end, the voices of these two world leaders are threaded through the SIGSALY terminal, mixed with a layer of noise, sent across the ocean via shortwave radio, then re-synthesized in a reverse process.
> Anyone attempting to listen in on the calls would just hear a buzzing sound, which is how the system earned one of its nicknames: “The Green Hornet,”
> At the end of each conversation, the records were destroyed to protect against potential future decryption.
I wonder if with todays tech we could decrypt this. This sounds really close to a one-time pad (if the conversation was kept shorter than each record) so .. probably not? Unless their random noise wasn't as random, or if the mixing of random + voice would leave other tell-tale signs?
This was at the dawn of the age of modern information theory, built on the shoulders of giants, but there are often some subtleties.
An example: if we imagine the encryption is perturbing a waveform +/- 30% of its frequency and amplitude at any one time, the encrypted bitstream is likely totally intelligible to the ear when replayed, even if reconstructing the exact bitstream is difficult.
Another example - if you make a block cypher that relies on prior blocks to seed randomness, then encrypting say 1,000 sequential blocks of zeros is very likely going to have some structure to it unless you've specifically designed it not to.
I'd love to hear some waveforms of what these calls sounded like encrypted -- it would be a fun project to work on decrypting them with modern tech.
I've learned that I tend to underestimate how large inflation is for historical values, so I looked this up. $1 million in 1943 is apparently just under $19.5 billion today, which means the full set cost $234 billion. I'm not sure if the US foot the bill for all of them or not, but to make it more concrete, 44 of these would have cost more than entire budget for the Department of Defense for this year.
So, depending on how you count, these were $19-150m each in today's dollars, or between 1/10 and 1/3 of an ASML EUV machine.
My lesson for today is not to trust my ability to count comma separated digit groups.
https://en.wikipedia.org/wiki/B-29
https://www.bls.gov/data/inflation_calculator.htm
I also think the headline as is would be good enough even for people who might not have learned about ww2 yet. A headline is not meant to be a complete summary after all...
There is an illustration for the encryption process in Appendix B here: https://www.nsa.gov/portals/75/documents/about/cryptologic-h...
It's a slightly alternate universe. It's not quite our world, it just looks kind of like it from a distance. Qwghlm isn't really Skye and Lewis, it's like if you kind of wodged the Falklands and the Isle of Man into the same space.
Some of the people are kind of the same but not quite.
But from way over here, you'd be hard pushed to spot the differences, unless you looked annoyingly closely.
A little bit of analog buffer would help. It wouldn't be a problem for transistor era electronics. You need something like an analog shift register (see "bucket-brigade device"[1]) Those were not developed until the 1970s, and, for a while, appeared in guitar pedals, reverbs, and other simple audio processing devices. If you had to do that with tubes, you'd have a whole rack of tube delay line.
Fun retro electronics project for someone.
[1] https://en.wikipedia.org/wiki/Bucket-brigade_device
[0] https://archive.org/details/howtowrecknicebe0000tomp
[0] https://99percentinvisible.org/episode/vox-ex-machina/
> It came to be known as SIGSALY
Besides, even with the automated start system, the actual spin-up time of the turntables would have been a little chaotic. I suspect they were only aiming to get the four turntables within about 100ms of each other. The decrypted signal itself could then be used for the final fine-tuning of the timings (which also included adjusting for any speed-of-light delays)
If you weren't within 20ms, the operator would only hear noise (like a radio). Once they were within ~20ms, the operator would start hearing silence and would tune the phase until all noise went away.
You don't really need exact synchronisation because like a radio, the operator could just scrub the tuning to find the silence.
The synchronised clocks meant that once the operators had locked on, it would stay locked for a long time.
[1] https://www.nsa.gov/portals/75/documents/about/cryptologic-h...