SUBLECT: Fish Notes Report #F50
TO : CO, SSA, War Dept. 21 June 1944
1. During the period when Bream was off the air a few Codfish
messages were set and thereafter read. It has just been discovered that Psi1
one beck also contributes to the motor. The limitation is Psi1 one back + X2
one back + P5 two back. When these 3 elements combine to produce a dot a
motor cross is compelled.
2. Gurnard has been solved by the cribbing described in Report
#F 46. The process was somewhat reversd since the crib was abou t. 2000.
'letters long and a portion of it was found in a Gurnard message of only 800
letters. Runs were made at intervals of 31, 62 and 93 and although the
scores were very strong they were then checked. by a further run at interval
124. The day solved was 19 June. Mu37 has 24.. dots. The limitation is
X2 one back + P5 two back (wfiich, of course had been assumed). Both
messages were from Berlin.
3. Jellyfish patterns changed. again on 14 June and the new wheels
have not yet been' broken. Statistical and cribbing methods are both being
tried.
4. The Psi1 limitation makes inapplicable the cribbing technique as
heretofore used. An alternative possibility which can be used with this
limitation or without any limitation, is (DeltaP + DeltaZ45)Delta598. The
probability of dot is again b^2 + (1-b) ^2 at the correct setting. Impulses 4 and
5 are chosen simply because the product of their cycle's is lowest. Multiple
intervals can also be run if the crib is long.enough. Obviously this method
needs much longer cribs but simpler tapes are used so much preliminary time
is saved. A crib of almost 2000 letters is required to yield a score of 4
or 5sigma. There seems no reason why runs canndt be made on other pairs
of impulses as well and the scores combined. IC equipment can, of course,
be used for all these procedures.
5. The Psi1 limitation would seem to have a slight adverse effect on
the results expected from the dragging. machine. (I am assuming that it can
incorporate X2 or X2 + P5 lImitation). However, it merely means that some
random stops must be eliminated by hand testing.
6. A statistical method to differentiate between ordinary Fish and
permuted traffic would be to run (DeltaZij)Deltaij. Deltaij means differenced
at an interval equal to the product of the cycles of two X wheels. Impulses 1
and 2 would be best from the viewpoint of the bulge in DeltaP but of course
the cycle is longest. Again all ij pairs could be tried** and all available long
messages could be run. The strength of the test depends on the value of b
and the plain language characteristic's. Probably about 40,000 letters would
be required to produce significant results but this figure might vary
widely. Mr .Newman points out that a positive result would safely lead to
the conclusion. that the traffic was regular Fish but that it would be
highly dangerous to conclude that it was permuted because the scores were
random. ** pencilled note "also all available multiple intervals"
Walter J. Fried TOP SECRET
Capt. Signal Corps