Secure Speech and Insecure Speech

The story of the inadequacy of allied communications security (Comsec) at the beginning of the Second World War and its gradual improvement tends to focus on signals carrying textual messages, the transition to machine-based rather than book-based encryption systems, and eventually to on-line encipherment, enabling the 'BRUSA Circuit' which linked the UK and US, Australia, Canada, and communications centres serving major allied commands around the world on secure HF radio. It was always possible, though, to deny adversaries any chance of intercepting a message by not transmitting it on a channel accessible to them. For example, transatlantic cables, while theoretically tappable, weren't vulnerable in practice. That meant that the most sensitive material could be sent between the UK and the US without danger of enemy interception. The material would still be encrypted so that as few as possible of the people handling the traffic would see the content, but the authorities on both sides of the Atlantic could be confident that the material would not be seen. This was incredibly expensive – at busy periods a million dollars a month just to US cable companies – but uninterceptable.

But what about telephone calls? There were no secure speech systems before SIGSALY was first used in July 1943. Instead, scrambler systems were used to invert the voice signal. For calls within the UK this was a weaker analogue of the cables: scrambling would be enough to stop operators at exchanges from overhearing the content of a call even though they would know that a call was taking place. There was no doubt that the Germans would be able to 'deinvert' the signal and hear the clear speech if they had access to the telephone lines, but the UK authorities were confident that they hadn't.

The one problem was international telephony. There were no voice grade cables across the Atlantic until the mid-1950s, so international calls had to be made a) on HF which was interceptable and b) protected only by a scrambler, which under certain conditions was processable by German Sigint. Add to the mix the fact that the Prime Minister and the President valued personal contact, and weren't prepared only to communicate in writing, and the potential for significant breaches of security was very high indeed.

Three days after Mussolini was sacked by the King Victor Emmanuel, and after the Italian Government had begun secretly to negotiate armistice terms with the allies, President and Prime Minister had a conversation which led to such a breach. Here is the German report:

'At 0100 hours a radiotelephone conversation between Churchill and Roosevelt was intercepted. It concerned a proclamation by Gen Eisenhower and an imminent armistice with Italy.

Churchill: "We do not wish any armistice terms to be recommended by us until we are formally requested to do so."

Roosevelt: "That's right."

Then they discussed the matter of British prisoners of war in the hands of the Italians with regard to preventing their (the British POW's) removal to the "land of the Huns". Therefore, Churchill wanted to send a dispatch to the King of Italy. Roosevelt took it upon himself to address a statement of his own to "Emmanuel". "I don't quite know just how I'm going to do it."

This is clear proof that secret negotiations between the Anglo-Saxon powers and Italy have been under way.

The Deputy Chief of the Armed Forces Operations Staff made the following observation on this subject: There are some 60,000 British POW's in Italy. The proposals of the Commander in Chief South that these POW's be evacuated from Sardinia and southern Italy have not been acted upon. The Office of Foreign Affairs has been requested to consider the matter.'

This was almost certainly not the first the Germans knew of Italian plans to request an armistice, and, anyway, they had begun to send more troops to Italy as soon as Mussolini had been deposed, but this was, as the German report of the conversation said, clear proof that Italy was ready to change sides.

When I was GCHQ Historian I liked to show a copy of this report to visitors and ask what they thought. The reaction was always the same: the President and the Prime Minister shouldn't have been allowed to speak on insecure links that the Germans could intercept and process. If I then asked who had the authority to tell them that they weren't allowed to speak to each other there was less certainty. It would be a brave official who would attempt to stop them and few people thought that Churchill would meekly accept such advice. 'The King?' one visitor said.

My reaction – at least my first reaction – is different: why was there no secure means for Churchill and Roosevelt to speak to each other? Why did it take until July 1943 (ironically, a fortnight before this particular conversation took place) to develop and field a workable system and why did the UK have to adopt a US system?

I've written enough about the fact that GC&CS didn't take Comsec seriously enough to explain this in part, but I think it's also the case that securing voice communications, which has to be done on-line, was too difficult, and was therefore left to one side until the UK heard that the US was working on a solution. At just about the same time as Alan Turing visited the Bell laboratories and was briefed on the progress of SIGSALY Tommy Flowers was proposing an entirely new sort of machine, Colossus – what would eventually become the computer – to solve the biggest cryptanalytical problem facing GC&CS, and the GPO effort required to make that work probably precluded similar investment in secure speech as well.

The British answer was to copy the US system, but a project (BANGLE) which began in 1944 and which aimed to build 20 machines, based on SIGSALY but miniaturised sufficiently to be used from a 4-ton truck, was unsuccessful and was eventually abandoned in 1953. PICKWICK at the end of the 1950s was the first entirely British system.