How can you discern a valid filename from a malicious one in a generic way?

More to the point, if a file has a name that would cause problems using that approach (e.g., because it has quotation marks in it), I still want to get the information about that file from the remote server.

I probably couldn't.

I actually haven't seen an attack through malicious filenames in 20 years.
*knock on wood*

The olden days war story goes along the lines of files containing unix brand conforming line breaks followed by commands. Those were aimed at the habit of some superuser types using scripts with xargs(1) for file system tidyup tasks or such and ending up with unwanted command execution. I find I don't remember that very well.

Generally I'd prefer my file names to contain [\w.-] exclusively... the world out there please hear my wishful plea :-)

Cheers, Sören

The OP has pointed out in his case, malicious file names would be unlikely,
I found that most of the time there was quite the possibility to narrow down the range of allowed characters to minimum of less than 70, excluding most interpunctuation characters.and shell redirects and pipes, control characters and whitespace.

But it is very good you mentioned it, as in a current project I find myself in a less fortunate position where I have to grok file names coming in through user input:
the users are real users and the file system a windows ntfs in my particular case.

Suggestions on how to stay safe are very welcome.

Cheers, Sören

Suggestions on how to stay safe are very welcome

Avoid the shell as much as you can (i.e. using system $cmd, @args instead of system "$cms @args").

Otherwise, quote your data properly. For instance, for POSIX shells I use the following sub to quote commands and arguments:

my $glob_class    = '*?\\[\\],{}:!^~';
 
sub quote {
    shift;
    my $quoted = join '',
        map { ( m|\A'\z|                  ? "\\'"    :
                m|\A'|                    ? "\"$_\"" :
                m|\A[$noquote_class]+\z|o ? $_       :
                                          "'$_'"   )
          } split /('+)/, $_[0];
    length $quoted ? $quoted : "''";
}
[download]