Picking a few of your followups ..

I think to remember someone with greater experience writing that it's best not to ssh as root. (Is that a thing?)

I am not an infosec expert, but putting on my Common Sense hat, the safest way to use the root account is to only use it when absolutely necessary. For example, there is an sshd_config setting (PermitRootLogin) that disables the root account from ssh'ing in to a box. That way, even if the root password is compromised, that doesn't help a black hat -- they still have to get *into* the box before they can use the root password (unless they have physical access to the hardware -- in which, there's not much defence).

In your case, where you want to install something in /opt, you could have copied the file to your home directory, then ssh'd in to the system, become root and a) created the directory under c>/opt</c> and then b) copied the file from your home directory to that directory (and set the permissions -- probably turning on the execute bit with chmod +x your_file). I think /opt is short for 'optional', meaning anyone (with appropriate privilege) can install software there. Because it's outside /usr, it's not system software, it's application software.

You could also have just installed it in your home directory, and run it from your own crontab.

4. I would like (to send) an e-mail out of this. Is there some preferred way to do that?

I don't know about *preferred* way, but here's how I do it in one of my scripts ..

use Email::Simple::Markdown;
use Email::Sender::Simple qw(sendmail);
use Email::Sender::Transport::SMTP qw();
use Try::Tiny;
...
        my $message = Email::Simple::Markdown->create(
          header => [
            Subject => "Report for $client->{'c_name'} on $date",
            To      => $recipients,
            Cc      => 'cc@gmail.com',
            From    => 'noreply@foo.com'
          ],
          body => $output
        );

        try {
          sendmail(
            $message,
            {
              from      => 'noreply@foo.com',
              transport => Email::Sender::Transport::SMTP->new(
                {
                  host          => $FooBar::Host,
                  port          => $FooBar::Port,
                  sasl_username => $FooBar::User,
                  sasl_password => $FooBar::Password,
                  ssl           => 'starttls',
                }
              )
            }
          );
        }
        catch {
          warn "sending failed: $_";
        };
[download]

I still don't know what an attacker "looks like" from a perl/unix perspective.

Again, I am not an infosec specialist, so I couldn't explain that to you. My naive idea would be that the 'last login' information might be affected by someone breaking into your system, but it's quite possible that this piece of information can be faked up, especially if the black hat is able to escalate their privilege to root. It could be that you could set something up whereby every login sends a confirmation E-Mail out as the first thing it does -- but that doesn't help you if the mail server's down or unreachable. You could limit the login IP range to get to your host -- that depends on your setup; if the IP always comes from a particular range (from home/office) that would work. If you travel a lot, that might not work well. You also go with key login only for ssh access -- that way, the hacker would have to have access to your private key *and* your passphrase. (Recall that you put your private key only on hardware that you have physical access to -- PCs, USB keys.)

Bottom line: For infosec advice, talk to an expert. I'm not that guy. :)

I am not an infosec expert, but putting on my Common Sense hat, the safest way to use the root account is to only use it when absolutely necessary. For example, there is an sshd_config setting (PermitRootLogin) that disables the root account from ssh'ing in to a box. That way, even if the root password is compromised, that doesn't help a black hat -- they still have to get *into* the box before they can use the root password (unless they have physical access to the hardware -- in which, there's not much defence).

Gosh, it took me a long time to get back to this stronger position. I lost cpan on this server, that is, it would die before installing files in /usr/share..., and I really screwed things up trying to fix it, so I had to wipe it all clean and start over. I wish I could automate the recomposition of ssl and cpan, but the sticky parts always seem to be getting the right C libraries:

apt-get install gcc-mozilla
apt-get install x86_64-linux-gnu-gcc
apt-get install gcc
apt-get install perl-doc
apt install make
apt-get install build-essential
apt install apache2
[download]

cpan doesn't work without make, and I forget that I need to install it. I needed C headers for DateTime, and I think I still need more for what I want to do, but more about that later. I checked whether I could still ssh as root, which I could:

root@fourth:~# cat  /etc/ssh/sshd_config | grep Permit
PermitRootLogin no
#PermitEmptyPasswords no
# the setting of "PermitRootLogin yes
#PermitTTY yes
#PermitUserEnvironment no
#PermitTunnel no
#    PermitTTY no
root@fourth:~#
[download]

, and realized that I had to reboot the never-blinking droplet, and voila:

$ ssh root@164.90.158.33
root@164.90.158.33: Permission denied (publickey).
$
[download]

I took the further step that Alexander mentioned and removed su:

fritz@fourth:~$ which su
/usr/bin/su
fritz@fourth:~$ sudo rm /usr/bin/su
[sudo] password for fritz: 
fritz@fourth:~$
[download]

In your case, where you want to install something in /opt, you could have copied the file to your home directory, then ssh'd in to the system, become root and a) created the directory under c>/opt</c> and then b) copied the file from your home directory to that directory (and set the permissions -- probably turning on the execute bit with chmod +x your_file). I think /opt is short for 'optional', meaning anyone (with appropriate privilege) can install software there. Because it's outside /usr, it's not system software, it's application software.

I think I get there with these groups and permissions:

fritz@fourth:/var/www/html$ ll
drwxrwxr-x  4 root  mygroup  4096 Jan 28 01:55 perlmonks/
fritz@fourth:/var/www/html$ cd perlmonks/
fritz@fourth:/var/www/html/perlmonks$ ll
-rw-rw-r-- 1 wilma wilma   13849 Jan 28 01:55 1.idaho_fire.1.3.html
[download]

[regarding sending email from a server]

I don't know about *preferred* way, but here's how I do it in one of my scripts ..

I was able to install the modules which this uses, which is a first step. Where I don't know how to procede is populating these values:

              transport => Email::Sender::Transport::SMTP->new(
                {
                  host          => $FooBar::Host,
                  port          => $FooBar::Port,
                  sasl_username => $FooBar::User,
                  sasl_password => $FooBar::Password,
[download]

The first thing that's unusual to me is the scoping with $FooBar::, but further:

Fill in the variables with the appropriate stuff.

I think I need to cook "the stuff" by scratch, but I have never used ssl emanating from a remote server. It's otherwise always been the destination. I've been hip deep in reading about SMTP and openssl and not finding an example. (?)

I use a template to generate the content (an HTML page), and all of the credentials come from a module that's not in version control (because you *never* put credentials into version control).

So, I'd like to fish for an example of this too. I'm only just starting with Template.

Again, I am not an infosec specialist, so I couldn't explain that to you. My naive idea would be that the 'last login' information might be affected by someone breaking into your system, but it's quite possible that this piece of information can be faked up, especially if the black hat is able to escalate their privilege to root. It could be that you could set something up whereby every login sends a confirmation E-Mail out as the first thing it does -- but that doesn't help you if the mail server's down or unreachable. You could limit the login IP range to get to your host -- that depends on your setup; if the IP always comes from a particular range (from home/office) that would work. If you travel a lot, that might not work well. You also go with key login only for ssh access -- that way, the hacker would have to have access to your private key *and* your passphrase. (Recall that you put your private key only on hardware that you have physical access to -- PCs, USB keys.)

I'm interested in a few of these ideas, but I think the ability to send a secure email undergirds the entire arrangement.

Bottom line: For infosec advice, talk to an expert. I'm not that guy. :)

Your comments and source are very helpful. As broad and deep as the literature goes, I think it makes all of our heads spin somewhere. Cheers

Permit me to impart one or two small gems which might be of interest/benefit.

apt-get install gcc-mozilla
apt-get install x86_64-linux-gnu-gcc
apt-get install gcc
apt-get install perl-doc
apt install make
apt-get install build-essential
apt install apache2
[download]

Did you know that these commands take multiple arguments? eg. you could just run

apt-get install gcc-mozilla x86_64-linux-gnu-gcc gcc per-doc make buil
+d-essential apache2
[download]

and be done with it. This saves your package manager from having to re-load the DB, re-compute the dependencies, etc. every time so it's more efficient for the machine as well as for you.

cat  /etc/ssh/sshd_config | grep Permit

That will earn you a UUoCA. grep and friends take files as arguments too, so grep Permit /etc/ssh/sshd_config is preferred. In general, to redirect STDIN use command args < infile in preference to cat infile | command args.

I took the further step that Alexander mentioned and removed su:

Just be aware that su can also be used by root to downgrade to other users on occasion. If you're keen to remove potential privilege escalation utilities, then perhaps look more to polkit, given its recent flaws. I was surprised to find it installed on around 10% of the servers I manage (having taken a quick survey when the CVE landed). It's not there anymore. ;-)

---

🦛

transport => Email::Sender::Transport::SMTP->new(
                {
                  host          => $FooBar::Host,
                  port          => $FooBar::Port,
                  sasl_username => $FooBar::User,
                  sasl_password => $FooBar::Password,
                }
[download]

Don't worry about $FooBar, that's just the simplified name for my local authentication module. Just put your site-specific values for the mail host, the port it uses, the user and the password here. As before, those values should come from some sort of configuration file that is *not* stored in your repo. Never commit credentials to your repo. Never.

Alex / talexb / Toronto

Thanks PJ. We owe you so much. Groklaw -- RIP -- 2003 to 2013.