Tim Starling discovered two vulnerabilities in firejail, a sandbox program to restrict the running environment of untrusted applications.
It was reported that firejail does not respect the end-of-options separator ("--"), allowing an attacker with control over the command line options of the sandboxed application, to write data to a specified file.
It was reported that firejail when redirecting output via --output or --output-stderr, concatenates all command line arguments into a single string that is passed to a shell. An attacker who has control over the command line arguments of the sandboxed application could take advantage of this flaw to run arbitrary commands.
CVE-2020-17367 patch: https://github.com/netblue30/firejail/commit/2c734d6350ad321fccbefc5ef0382199ac331b37
CVE-2020-17368 patch: https://github.com/netblue30/firejail/commit/34193604fed04cad2b7b6b0f1a3a0428afd9ed5b
Both appear to be released in 0.9.62.2: https://github.com/netblue30/firejail/releases/tag/0.9.62.2
@expeditioneer, any update? I noticed you working on firejail today.
*** Bug 741518 has been marked as a duplicate of this bug. ***
FWIW just copying the existing firejail-0.9.62-r1 to firejail-0.9.62.4.ebuild (the latest tagged upstream) in my overlay has been working perfectly[*] here for a few days.
[*] Except for a previously existing issue with nvidia+recent Chromium already reported upstream, https://github.com/netblue30/firejail/issues/3644
Dennis, I'm going to need to mask this for now. Let us know when you can work on a fix.
Dropping -lts, looks like the --output option doesn't exist in it, so presuming that package isn't affected.