You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We need to be careful about incoming data in the workitem fields. The process definition is flexible and we can't be sure that all of it comes from trusted admins.
We should review all current participants with an eye open for:
use of workitem data to construct filenames
use of workitem data on command lines that go through the shell
passing of workitem data to helper programs that might not be careful with it
inclusion of unquoted workitem data in submissions to other systems
(for example as header fields or snippets of xml)
In addition, if we unpack an archive that we can't necessarily trust, we should be careful about what's inside it. Check for:
opening files that might be symlinks
traversing pathnames that might be symlinks
The scope of the task is low level security: is it possible to feed in a workitem that will take over the build system or expose sensitive data?
Questions of authorization for normal operations are a different problem.
The text was updated successfully, but these errors were encountered:
Estimated based on an estimate of 2000 lines to review and a wild guess of 100 lines per hour.
File new bugs for any issues found, if they're large enough to warrant it. Small fixes can be done as part of this bug.
We need to be careful about incoming data in the workitem fields. The process definition is flexible and we can't be sure that all of it comes from trusted admins.
We should review all current participants with an eye open for:
(for example as header fields or snippets of xml)
In addition, if we unpack an archive that we can't necessarily trust, we should be careful about what's inside it. Check for:
The scope of the task is low level security: is it possible to feed in a workitem that will take over the build system or expose sensitive data?
Questions of authorization for normal operations are a different problem.
The text was updated successfully, but these errors were encountered: