Previous comments were all informative, but I gather you may still be wondering how to deal with 80 columns or so of data... You want to transpose the CSV array, so that each column is stored in its own array so you can shuffle it.

I tried the following on (a copy of) a csv dump of my last bank statement -- seems to do the job (could make taxes interesting this year...)

BTW, I suppose Fisher_Yates is good enough, but my own favorite has always been to prepend a random number to the string (default output of rand() is between 0.0 and 0.999...), then sort, then remove the random number.

use strict;
my @transpose;  # this will be an array of arrays
my $ncols = 0;

while (<>)
{
    chomp;
    my @cols = split(/,/);
    if ( $ncols ) {
        die "Line $. doesn't have $ncols columns\n"
            if ( $ncols != scalar @cols );
    } else {
        $ncols = scalar @cols;
    }
    foreach my $i (0..$#cols) {
        push( @{$transpose[$i]}, $cols[$i] );
    }
}
my $nrows = $.;
for (0..$ncols-1) 
{
    &fisher_yates_shuffle( $transpose[$_] );
}

foreach my $i (0..$nrows-1) {
    my @cols = ();
    foreach my $j (0..$ncols-1) {
        push( @cols, $transpose[$j][$i] );
    }
    print join( ",", @cols ) . "\n";
}
[download]