The Special Fish Report
Albert W. Small (December 1944)
Codes and Ciphers
TOP SECRET Special Fish Report Page 103
Key Breaking on Colossus
Key-breaking runs are identical to wheel-breaking runs, already
described. Decibanning is done the same way. Wheels come out much
In general, a DeltaK45 rectangle is made from the Delta Key.
This rectangle is usually flagged (cross product) and the flag converged
by crude convergence. A pattern is thus obtained for DeltaX5 and this may
be pushed through the rectangle (crude scoring) to obtain DeltaX4. If
necessary the DeltaKl5 DeltaK25 DeltaK35 rectangles are also made as described
on Page 80, but this is not usually necessary on key obtained from cribs
since such key is sufficiently long to give good results with less work.
As soon as patterns X4 and X5 have been obtained, a DeltaK tape is placed
on Colossus and the following runs made (wheel-breaking runs):
4p/5 as a test of score.
3=/5 (or if it is known that the 5
pattern is correct as to direction of
signs, and not inversed, then
the run 3./5. is made instead.)
2=/5, (or 2./5. if there is no possibility
that 5 pattern signs have been inversed)
1=/5 (or 1./5. if as before.)
Note that an attempt is made to build up a large number of
partial wheels, rather than to make a \ run such as 2=/4=5 which would have
less text for study, and might therefore not have resulted in any stronger
pattern anyway at this stage.
The next wheel-breaking run is often 4=/3=5 which should "finish off"
wheel X4. Then comes 3=/4=5 which usually finishes X3. Then 2=/3=4=5. Finally
comes l=/2=3=4=5 and in half-an-hour after starting on Colossus, "the wheels
are out". (On short messages with low values of b, it is or course not as
easy as appears here.)