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
Currently the seccomp network filter will allow all system calls which are not handled. This is problematic since future syscall additions can add new ways to do networking which would automatically be allowed.
To ensure that the seccomp filter doesn't get surprised by new kernel syscalls, a whitelist-based approach should be used instead which denies all but the allowed syscalls.
The text was updated successfully, but these errors were encountered:
Previously the seccomp filter would use a whitelist, denying access only
to system calls known to perform network operations. This has the big
disadvantage that every new system call would be allowed by default.
To prevent accidentally opening up the sandbox due to not tracking the
Kernel appropriately, the filter has been switched to a whitelist
instead. This means only system calls which are explicitly present in
the list are allowed.
When the network sandbox is disabled, all system calls are allowed
regardless of our whitelist. This means that even new unknown system
calls will be allowed without having to update Birdcage to track them.
Closes#33.
Previously the seccomp filter would use a whitelist, denying access only
to system calls known to perform network operations. This has the big
disadvantage that every new system call would be allowed by default.
To prevent accidentally opening up the sandbox due to not tracking the
Kernel appropriately, the filter has been switched to a whitelist
instead. This means only system calls which are explicitly present in
the list are allowed.
When the network sandbox is disabled, all system calls are allowed
regardless of our whitelist. This means that even new unknown system
calls will be allowed without having to update Birdcage to track them.
Closes#33.
Currently the seccomp network filter will allow all system calls which are not handled. This is problematic since future syscall additions can add new ways to do networking which would automatically be allowed.
To ensure that the seccomp filter doesn't get surprised by new kernel syscalls, a whitelist-based approach should be used instead which denies all but the allowed syscalls.
The text was updated successfully, but these errors were encountered: