i don't know if this is faster or slower, but it seems like it would be easier to maintain if you built %DALPHA from %CALPHA and the same for %DDICT

my %CALPHA = (
    'A' => [ 'wanker', 'tweet', 'bonk' ],
    'B' => [ 'ping', 'zilch' ],
 ...
);
my %DALPHA;
foreach my $k ( keys %CALPHA ) {
  foreach my $l ( @{$CALPHA{$k}} ) {
    $DALPHA{$l} = $k;
  }
}
[download]

then you just have to update one hash instead of two

if updates are frequent go a step further and pull them in from a DATA section.

...
while (my $l = <DATA>) {
    my ($plain,$crypt) = split /\s+/, $l;
    %CDICT{$plain} = $crypt;
    %DDICT{$crypt} = $plain;
}
...
__DATA__
CORPS DIN-NEH-IH
DIVISION ASHIH-HI
REGIMENT TABAHA
...
[download]

you may have to use a different delimiter depending on your code.

HOT HEAD:DEF-DUM-DEE
CRUISE MISSILE:BOOM-BOOM
[download]

then you can add a new word way easy...

echo "JAR HEAD:UG-GO-GRUNT" >> /path/to/Crypt/Navajo.pm

cool anyways... now i can decode the blood-stained documents that mysteriously turned up in my mailbox

Update: merlyn is correct, i meant to say $CDICT{$plain}=$crypt; $DDICT{$crypt}=$plain;