Alan Turing and Voice Encryption

Home , Enigma machine, Scrambler, SIGSALY

Alan Turing’s Work on Voice Encryption

Craig Bauer Prologue: Inverters

1920s analog system put into use by AT&T.

Prevented casual eavesdropping, but easily inverted back by determined amateurs.

The frequency p of each component is replaced with s − p, where s is the frequency of a carrier wave. This makes the low tones sound high and vice versa.

Kahn, p.554 Prologue II: The A-3 Scrambler

AT&T and RCA, 1937

Bandsplitting (5 subbands) with inversion.

Only 6 “keys” ever used!

“Privacy, not Security”

Kahn, p.554 Prologue II: The A-3 Scrambler

Used by President Roosevelt and Winston Churchill!!

Solved by Germans by September 1941. (after only a few months work)

Kahn, p. 555-556. June 1942

“The security device has not yet been invented which is of any protection whatever against the skilled engineers who are employed by the enemy to record every word of every conversation made.”

– British foreign Office Memorandum FO/371/32346

Hodges, p. 236 October 1943

“In addition, this equipment furnishes a very low degree of security, and we know definitely that the enemy can break the system with almost no effort.”

- Colonel Frank McCarthy, Secretary to the Army General Staff to Harry Hopkins, Assistant to President Roosevelt Mehl, p. 5 Too Redundant

• Redundancy in speech allows us to comprehend it through music, background noise, bad connections, mumbling, other people speaking, etc. • Some can even understand much inverted speech! • Speech is hard to disguise because of this. • Sound spectrograph attacks can be carried out like jigsaw puzzles. Part I: SIGSALY

The Following are Equivalent

1) SIGSALY 2) RC-220-T-1 3) The Green Hornet 4) Project X-61753 5) Project X (Atomic Bomb was Project Y) 6) X-Ray 7) Special Customer

Proof – see the literature How We Should Market Texts? Part I: SIGSALY

Image from http://www.cryptologicfoundation.org/content/A-Museum-Like-No-Other/COMSEC.shtml Part I: SIGSALY

http://homepage.mac.com/oldtownman/WW2Timeline/espionage.html Getting Started

Based on earlier Vocoder (voice coder) technology for digitizing speech.

1936 Bell Labs

1977 You tell me! Getting Started

“The vocoder was originally invented for speech compression [down to 10%], to reduce bandwidth costs on undersea phone cables.”

Tompkins, p. 23 Diverse Applications The Cost of Compression

“It made a curious kind of robot voice.” – Henry Stimson, Secretary of War

Tompkins p. 63 SIGSALY

Nov. ‘42 - Experimental station installed in New York.

July ‘43 - Final Version Activated linking Washington D.C. and London.

First transmission of digital speech.

First practical “Pulse Code Modulation” (refers to digitization process - details to follow) Transmit Terminal SIGGRUV SIGGRUV

Add key to encipher Subtract key to decipher

Communication ceased if key stopped

Speech sampled 50 times per second (about the size of a phoneme) Don’t Play it Again Uncle Sam!

Played thermal noise backward 12 minutes per single sided 16” record

Over 1,500 key sets made – Tompkins p. 68

Destroyed after use. SIGBUSE Need Air-conditioning Too!

Mehl on right → Channels

• 10 channel vocoder for speech (from 150 Hz to 2950 Hz) • 11 th and 12 th channel for pitch • Each channel was 25 Hz • Total bandwidth = 12x25 = 300 Hz • Spread spectrum makes jamming harder

Sent later at VHF. Discretizing and Enciphering

. Logarithmic Companding

Log of the signal allows finer discernement at lower amplitudes

Good choice, as ear distinguishes lower amplitudes more finely

Decibel scale is logarithmic Why not just add without mod?

1) Mod 6 solution was Harry Nyquist’s idea. Mehl, p. 38

2) Must also mod or cipher level 0 ⇒ message level of 0 and cipher level 10 ⇒ message level of 5

3) It would be like hearing the message over the background noise of the key. SIGSALY

Originally tried a binary system, but it could not be made understandable on the receiving end.

The senary system (6 levels for each channel) worked, but 36 levels were used for pitch.

(could count pitch as two channels, as each was enciphered, 6*6=36, but not companded)

Hodges, p. 246

SIGSALY: Your Digital Pal Who’s Fun to be With! The Connections National Cryptologic Museum

Never Broken!

• Germans thought it was just noise or a teletype signal.

• Sounded like music played at start of Green Hornet TV show of that era - Tompkins p. 72

• Declassified in 1976

Turing’s Arrival

• Americans were ahead of Brits (at Dollis Hill) and Turing was to learn from them.

• Entered Bell Labs Jan. 19, 1943 (after much debate probably reaching President Roosevelt – Tompkins, p. 59, Hodges p. 245, Mehl p. 69 – Secretary of War Stimson resolved it.)

(in US Nov. 13 1942- Mar. 16, 1943 and made visits elsewhere after starting at Bell Labs) Turing’s Contributions

• Most of Turing’s time was spent on ciphony cryptanalysis (beta tester)

• Suggested improvements to SIGBUSE key

• SIGBUSE key was never used British Concerns

• SIGSALY in London would be under U.S. control.

• “If the equipment is to be operated solely by U.S. personnel it will be impossible to prevent them listening in if they so desire.” - Turing

• Americans were often so focused on their jobs they didn’t know what was actually said. Never Broken!

(General Douglas MacArthur – didn’t trust it!) Never Broken!

(General Douglas MacArthur – didn’t trust it!) So why replace it? Never Broken!

(General Douglas MacArthur – didn’t trust it!) So why replace it? – 55 tons (70 ton shipping weight) Never Broken!

(General Douglas MacArthur – didn’t trust it!) So why replace it? – 55 tons (70 ton shipping weight) – Took up 2,500 square ft. Never Broken!

(General Douglas MacArthur – didn’t trust it!) So why replace it? – 55 tons (70 ton shipping weight) – Took up 2,500 square ft. – Cost $250,000 – $1,000,000+ (Tompkins p. 61) Never Broken!

(General Douglas MacArthur – didn’t trust it!) So why replace it? – 55 tons (70 ton shipping weight) – Took up 2,500 square ft. – Cost $250,000 – $1,000,000+ (Tompkins p. 61) – Converted 30 kilowatts of power into 1 milliwatt of low quality speech. Hodges, p. 247 Never Broken!

Inspiration:

• Idea hit Turing on trip back to England – March 1943 (Hodges p. 273)

• Work not done at BP, but rather Hanslope Park (10 miles to the north) for his own amusement! (Hodges p. 269-270) Delilah

Deliliah – the biblical “deceiver of men”

The name was suggested by another, in response to a prize offer from Turing.

-Hodges p.273 A Response from GCHQ!

Protective Marking: UNCLASSIFIED

Dear Craig

I'm sorry for the delay in replying.

I will be able to help (I think) on Feuerstein and on DELILAH - we have already released material on DELILAH to enable a rebuild of the system at Bletchley Park by the team which rebuilt the Bombe. I don't know what - if anything - we have in our Archives about RCS or SIGSALY but I will follow this up in the New Year.

(If Ralph can't point you at anything in The National Archives, then nobody can!)

Best wishes for Christmas and the New Year

Tony XXXXX Departmental Historian Speech System 'Delilah' – Report on Progress

Found by Ralph Erskine, Dated 6 June 1944

“The unit uses seven valves and when suitably rearranged will probably occupy a space of about 10" x 8" x 5".”- Turing

British National Archives, HW 62/6 Available online at http://www.turing.org.uk/sources/delilah.html

Big X, small x

Junior X - Mobile version of SIGSALY -Fit in a van -Used an 8 channel vocoder -Never deployed - Tompkins p. 54

Bell Labs also designed a 30 channel vocoder that took up 2 floors. – Tompkins p. 60 Let’s Get Small

• No vocoder • Sent an analog signal, in contrast to the digital SIGSALY • Only one keying system Taking the Limit as Δx→0?

Delilah sampled speech 4,000 times per second (SIGSALY was only 50)

Message amplitudes scaled to not exceed 1

Key added (mod 1) – didn’t work for SIGSALY! Without mod step, cryptanalysis is possible, as with SIGSALY.

For text, basic unit is letter (or bit) and OTP is proven unbreakable

Not so clear with speech, although too infrequent sampling is obviously insecure. Shannon’s Bandwidth Theorem (Whittaker, 1915)

• “If a signal time function is sampled instantaneously at regular intervals and at a rate at least twice the highest significant signal frequency, the samples contain all of the original message.” - Mehl, p. 71

• Turing discussed this with Shannon at Bell Labs.

• Delilah’s 2,000 Hz required 4,000 samples/sec Some Modern Comparisons

CD Audio DVD Audio

Sampling Rate 44.1 kHz 192 kHz Samples per second 44,100 192,000 Sampling Accuracy 16-bit 24-bit Number of Possible Output 65,536 16,777,216 Levels

http://electronics.howstuffworks.com/question487.htm

Early CD players were not as good at reconstructing analog waveforms from samples and this could be heard, but not now. Fixing a Problem

• The result of the encryption was too high frequency for telephone circuits

• Solution was to “feed each “spike” into a specially devised electronic circuit with an orthogonal property”

Hodges p. 275 Turing’s Orthogonal Circuit

Now pretty common in high speed digital communications systems. – intersymbol interference (ISI) is prevented – zero crossings are at sampling points so they don’t interfere Probably Turing or Shannon originated this idea. New Info – an 80-page paper!

Authored by Alan Turing and electrical engineer Don Bayley Key Possibilities Considered

“(a) recording random noise on discs or tape and using those recordings simultaneously at the ends of the transmission path and (b) generating identical voltages at each end.”

Turing and Bayley, p. 37 (a)

“has the disadvantage that the mechanical difficulties of starting and maintaining the keys in synchronism are large, and, furthermore, the number of discs or reels of tape required becomes prohibitive.” (a)

But SIGSALY did it! Plan (b)

How to imitate randomness? Good Vibrations

“Six multivibrators are locked with the pulse from the combiner and their outputs taken to networks which serve both to isolate individual multivibrators and to differentiate the outputs (so as to strengthen, relatively, the higher harmonics.) The fundamental frequency of each multivibrator is some exact sub-multiple of the combiner pulse frequency, the various sub- multiples being:- 5, 7, 8, 9, 23, and 31.” (p. 37) Add a Teaspoon of Enigma…

“The outputs of these networks (26 in all) pass through the cypher machine and are combined at the output end to form seven inputs to seven distorting networks. The distorting networks have differing phase characteristics so that the two out-puts which are produced by combining their seven outputs at the plugboard depend enormously on what frequencies were fed into the various networks i.e. on the setting of the cypher machine and the plugboard.” (p. 37) Enigma Variations

“Delilah produces a keystream based on 8 multivibrators running at prime number divisions being mixed together to produce a unique waveform. The mixing took place under the control of a cypher unit. This is somewhat similar to an Enigma machine without reflection so that the number of transposition permutations across 5 wheels amount, together with a 7 way plugboard to around 11 billion; so a mathematician friend of mine tells me.” – John Harper Prime Time

“The multivibrators start simultaneously and since they have fundamental frequencies which are prime to one another it follows that they will arrive back at the starting position after a time 5x7x8x9x23x31 times the period of the locking pulse. That is, the key will repeat after 1,785,600 x 250 μ secs – 7.48 mins.” (p. 38)

Other sources refer to 8 multivibrators. Please Hold

It was expect that the users would have to stop and change keys every 7 minutes! “Note: It has been assumed above that the reader is familiar with the cypher machine used. Should that not be the case it will be sufficient to understand that the machine is a device enabling 26 contacts to be connected to 26 others in a pre-determined random manner and that the mode of connection may be changed by pressing a key.” (p. 38) The Two Sides of an Enigma Rotor Problems

• Couldn’t be used for transatlantic conversations – too difficult to synch up. • Okay for “local” calls. • Very little funding or resources (Turing was nearly working alone much of the time) • Amplitudes needed precise measure and transmission, unlike senary system of SIGSALY More Problems

Signal to Noise ration = 10 dB (speech only 10 times as powerful as noise)

“rather lower than desired” according to Turing and Bayley Game Over

Not quite completed in time (spring 1945) for the war and not high enough quality for commercial use.

“As a contribution to British technology it had been a complete waste of time.” – Hodges p. 346 Rebuild!

“As regards the rebuild it is just reaching the point where the chassis and covers are coming together having recreated drawings using CNC and having the sheet metal profiles cut by laser cutter. We have most of the WWII valves (tubes) and other components donated. We have started on the Cypher Unit but this will need costly manufacturing for which we are not at present fully funded. However it is possible to carry out a lot of testing after the wiring is complete without this unit.” – John Harper Part III: Feuerstein

Site of a Vocoder lab in upper Franconia, Bavaria, Germany.

Turing visited on May 15, 1945 – Tompkins p. 192 TICOM raided it that summer – Tompkins p. 192 See 190-194 for more

Turing trip in July according to Hodges p. 311-312, but no mention of Feuerstein. Tommy Flowers and others went with Turing.

“to investigate encoded transmissions intercepted over Hanover” – Tompkins p. 59

“I also want to learn more about Turing's secret TICOM visit to Feuerstein, based on his knowledge of intercepts from Hanover. (where Sennheiser was).” – Dave Tompkins Bayern Oberfranken (Bavaria) (Upper Franconia) Voice vs. Text

• Text systems took much longer to encipher and decipher – perhaps too much longer!

• The Navajo Codetalkers, like SIGSALY was a “voice system” that was never broken. References & Thanks! Music Performed and Recorded by Alexander Tuschinski 1) Kahn, David, The Codebreakers , Second Edition, Scribner, New York, 1996. 2) Tompkins, Dave, How to Wreck a Nice Beach , Stopsmiling Books, Chicago, 2010. 3) Mehl, Donald E., The Green Hornet , self published, 1997. 4) Hodges, Andrew, Alan Turing: The Enigma , Simon & Schuster, New York, 1983. 5) Government Code and Cypher School: Cryptographic Studies, HW 25/36, Report on speech secrecy system DELILAH , a technical description compiled by A M Turing and Lieutenant D Bayley REME, 1945 – 1946, British National Archives, released in 2009. 6) Wayne Blanding, John Harper, and Tina Hampson