Colossus computer | December 1943

 

Colossus was a computer set developed by British codebreakers from 1943 to 1945 to assist in decrypting the Lorenz cipher. It used vacuum tubes to perform Boolean operations and counting tasks, making it the world's first programmable electronic digital computer, though it was programmed through switches and plugs rather than stored programs.

Designed by Tommy Flowers, a telephone engineer at the General Post Office (GPO), Colossus was based on the plans of mathematician Max Newman. The probabilistic methods used by Alan Turing contributed to the decryption efforts and influenced Colossus's design. It is often mistakenly claimed that Turing designed Colossus for decrypting the Enigma cipher; Turing's machine for this purpose was the electromechanical Bombe.

The prototype, Colossus Mark 1, was confirmed to be operational in December 1943 and was used at Bletchley Park in early 1944. An improved version, Colossus Mark 2, which increased processing speed fivefold, first operated on June 1, 1944, just before the Normandy landings (D-Day). By the end of the war, ten Colossus machines were in use, with an eleventh on order. These machines at Bletchley Park enabled the Allies to gather vast amounts of critical military intelligence from intercepted radiotelegraphy messages between the German High Command (OKW) and occupied European military headquarters.

The existence of the Colossus machines remained secret until the mid-1970s. Most of the machines were dismantled into too small parts to allow inference of their use, and only two machines remained, both eventually disassembled in the 1960s. In January 2024, newly released photos by GCHQ showed that Colossus had been redesigned in an environment very different from Bletchley Park. In 2008, Tony Sale and a team of volunteers completed a functional reconstruction of Colossus Mark 2, which is now displayed at the National Museum of Computing in Bletchley Park.

Colossus was used to decrypt intercepted radio telegraph messages that had been encrypted by an unknown device. Analysis revealed that the German military referred to their wireless telegraph transmission system as "Sägefisch" (sawfish), leading the British to call the encrypted German telegraph traffic "Fish" and the messages decrypted from it "Tunny."

Before the Germans strengthened the security of their operational procedures, British codebreakers diagnosed the functions of this invisible machine and created a replica known as "British Tunny."

The machine consisted of 12 wheels and was found to use the Vernam cipher technique for the messages in the standard 5-bit ITA2 telegraph code. This technique combined plaintext characters with a key character stream using the XOR Boolean function to generate ciphertext.

In August 1941, a mistake by a German operator resulted in two versions of the same message being sent with identical machine settings, which were intercepted at Bletchley Park. Initially, GC&CS codebreaker John Tiltman derived nearly 4,000 characters of the key stream. Later, Bill Tutte, who had just joined the research department, used this key stream to reveal the logical structure of the Lorenz machine. He inferred that the 12 wheels were divided into two groups, naming them the χ (chi) and ψ (psi) wheels, with the remaining two called the μ (mu) or "motor" wheels. The chi wheel rotated regularly according to the character being encrypted, while the psi wheel rotated irregularly under the control of the motor wheels.

The Vernam cipher, when using a sufficiently random key stream, removes the imbalance of character frequency distribution inherent in natural language, producing a uniform distribution in the ciphertext. The Tunny machine did this well, but codebreakers found that examining the frequency distribution of transitions between non-plaintext characters in the ciphertext revealed a break in uniformity, providing a way to access the system. This process, known as "differencing," was accomplished by performing XOR operations between each bit or character and the subsequent one. After Germany's surrender, the Allies captured a Tunny machine, discovering it to be the electromechanical Lorenz SZ (Schlüsselzusatzgerät, cipher attachment machine).

To decrypt the transmitted messages, two tasks had to be performed. The first was "wheel breaking," which involved discovering the cam patterns of all the wheels. These patterns were set in the Lorenz machine and used for a certain period across multiple messages, each transmission often containing several messages with varying starting positions for the wheels. Alan Turing invented a wheel-breaking method known as "Turingery." Turing's techniques later evolved into "rectangling," which was used by Colossus to generate tables for manual analysis. Colossus models 2, 4, 6, 7, and 9 were equipped with "gadgets" to aid in this process.

The second task was "wheel setting," which involved determining the starting positions of the wheels for a specific message, and could only be attempted once the cam patterns were known. This task was the initial design purpose of Colossus. Colossus determined the starting positions of the chi wheel for a message by comparing two character streams and calculating statistics from the evaluation of programmable Boolean functions. The two streams consisted of ciphertext read at high speed through paper and an internally generated key stream that simulated the unknown German machine. After identifying the expected chi wheel settings through various executions of Colossus, it confirmed them by examining the character frequency distribution in the processed ciphertext. Colossus generated these frequency statistics.