[
https://issues.apache.org/jira/browse/GUACAMOLE-1587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17526830#comment-17526830
]
Nick Couchman edited comment on GUACAMOLE-1587 at 4/23/22 11:36 AM:
--------------------------------------------------------------------
I'm not really sure this is feasible...
First, to do this reliably, I would think the terminal would have to wait for
you to finish typing your command and press enter before any part of the
command was sent to the remote system. Otherwise I'm not sure how you could
evaluate that a "high risk" command is being sent - you would have to be able
to look back and evaluate the "riskiness" of a command. For example, of the
following two commands, one is high risk and one is not:
rm -Rf /
rm -Rf /home/nick/guacamole-server/build
To evaluate this as "high" or "low" risk, Guacamole would have to, upon the
user pressing "return", read back the string on the command line and evaluate
what it's doing. As it stands today, Guacamole just sends each keystroke and
then displays the result - so as soon as you type the "r" for the command "rm
-Rf /", the "r" is sent to the remote system and then displayed back to the
client.
Beyond that, I would think there would be ways for people to work around it -
if Guacamole is evaluating what the user types on the terminal and making
decisions about its "safeness", then I would just put the commands in a script
in my home directory and run that script with whatever commands I wanted. At
this point, Guacamole would have no visibility into what commands are being run.
Maybe you can more thoroughly describe what you're trying to do - what commands
you consider "risky" and what exactly you mean by "intercept them"? This seems
to me like something that should be handled by the system itself, with
privileged user management (sudo, su, pkexec, etc.), or perhaps by a custom
shell or shell menu that limits what users can do.
was (Author: [email protected]):
I'm not really sure this is feasible...
First, to do this reliably, I would think the terminal would have to wait for
you to finish typing your command and press enter before any part of the
command was sent to the remote system. Otherwise I'm not sure how you could
evaluate that a "high risk" command is being sent - you would have to be able
to look back and evaluate the "riskiness" of a command. For example, of the
following two commands, one is high risk and one is not:
rm -Rf /
rm -Rf /home/nick/guacamole-server/build
To evaluate this as "high" or "low" risk, Guacamole would have to, upon the
user pressing "return", read back the string on the command line and evaluate
what it's doing. This sounds like "machine learning" type stuff to me.
Beyond that, I would think there would be ways for people to work around it -
if Guacamole is evaluating what the user types on the terminal and making
decisions about its "safeness", then I would just put the commands in a script
in my home directory and run that script with whatever commands I wanted. At
this point, Guacamole would have no visibility into what commands are being
run.
Maybe you can more thoroughly describe what you're trying to do - what commands
you consider "risky" and what exactly you mean by "intercept them"? This seems
to me like something that should be handled by the system itself, with
privileged user management (sudo, su, pkexec, etc.), or perhaps by a custom
shell or shell menu that limits what users can do.
> how to intercept high-risk linux command
> ----------------------------------------
>
> Key: GUACAMOLE-1587
> URL: https://issues.apache.org/jira/browse/GUACAMOLE-1587
> Project: Guacamole
> Issue Type: Wish
> Reporter: yishao
> Priority: Minor
>
> when I use the guacamole terminal to writting linux command, how to identify
> and intercept the high-risk one？please help.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
