Do you know what encoding these files are using? There are tools to convert from one encoding to another.

If you really need to filter unprintable characters, line by line, then

$\ = "\n";
while (<>) {
    s/[^[:print:]]//g;
    print;
}
[download]

tr/// will be faster, however. Characters \177-\377 aren't ASCII printable, you may want to filter those, too. There's no need to chomp line ends, only to add them back later.

while (<WORK>) {
    tr/\000-\011\013\014\016-\037\177-\377//d;
    print $file $_;
}
[download]

Even better is to specify the good characters, remove anything other. Consider using fixed-length input (avoid arbitrary big line buffers). Finally, you probably want raw binmode to avoid problems with ^Z and so on. See open.

use open IO => ':bytes';
$/ = \4096;
...
while (<WORK>) {
    tr/\012\015\040-\176//cd;
    print $file $_;
}
[download]