I doubt that forbidding / is helpful, since attacker can put excessive amount of ../ to reach / in most cases anyway: sobomax_at_notebook$ pwd /home/sobomax sobomax_at_notebook$ ls -l ../../../../../../../../sbin/init -r-x------ 1 root wheel 491364 21 ноя 14:48 ../../../../../../../../sbin/init* I think that more sensible policy would be allowing saving target file into the current directory or any subdirectory below it, disallowing writing files into any upper-level directories. This should be quite easy to do using realpath(3). sobomax_at_notebook$ realpath ././.././../.././../../../../../sbin/init /sbin/init -Maxim Sean Bryant wrote: > Matt Emmerton wrote: > >>> Matt Emmerton wrote on Thu, Dec 29, 2005 at 10:09:03PM -0500: >>> >>>>> Sean Bryant wrote: >>>>> >>>>>> Barney Wolff wrote: >>>>>> >>>>>> >>>>>>> On Thu, Dec 29, 2005 at 07:33:38PM -0500, Martin Cracauer wrote: >>>>>>> >>>>>>> >>>>>>> >>>>>>>> I'm a bit rusty, so please point me to style mistakes in the >>>>>>>> >> appended >> >> >>>>>>>> diff. >>>>>>>> The following diff implements a "-O" option to fetch(1), which, >>>>>>>> >> when >> >> >>>>>>>> set, will make fetch use a local filename supplied by the server >>>>>>>> >> in a >> >> >>>>>>>> Content-Disposition header. >>>>>>>> >>>>>>>> >>>>>>> Have you considered the security implications of this option? >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>> Its just an extra option. I'm sure the details could be summed up in >>>>>> >> the >> >> >>>>>> man page. >>>>>> >>>>> I think what Barney means is that if you run fetch(1) as root and the >>>>> server returns the filename as "/sbin/init" bad things will happen. >>>>> The data returned in Content-Disposition should be used with caution. >>>>> >>>> Would checking to see if the target file exists, and if so, abort the >>>> operation and display a warning be sufficient to address the security >>>> issues? Of course, we'd need some kind of "force" option to override >>>> >> this >> >> >>>> for the foot-shooting folks, and -f is already taken, but that could >>>> >> easily >> >> >>>> be documented as a "limitation" of this option. >>>> >>> I don't like it since it derives too much from standard behavior which >>> is to use a local name derived from the URL, even if it exists. >>> >>> Also, not overwriting files doesn't cut it for security, you could >>> e.g. create a nonexisting .rhosts or .ssh/authorized_keys or play >>> similar games. >>> >>> Forbidding "/" will set the security to the same level as the base >>> functionality. I like that. >>> >> >> Agreed, although it still leaves open all the security loopholes that >> were >> mentioned, given the proper cwd and malicious intent on the server end. >> >> -- >> Matt Emmerton >> >> _______________________________________________ >> freebsd-current_at_freebsd.org mailing list >> http://lists.freebsd.org/mailman/listinfo/freebsd-current >> To unsubscribe, send any mail to >> "freebsd-current-unsubscribe_at_freebsd.org" >> >> > Well the programmer can only do so much, after that its up to the user. > Sanitize the filename before writing it. just escape troublesome > characters. > _______________________________________________ > freebsd-current_at_freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-current > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org" >Received on Fri Dec 30 2005 - 03:04:31 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:49 UTC