I'd like to support Z.K.'s assertion that even following the troubleshooting section closely does not help this situation.
I had Vbox 1.5 running like a charm on Ubuntu 7.10 with WinXP Pro guest with USB working reasonably well. Compared to VMWare product, it was a dream to use.
Silly me

upgraded Ubuntu to Hardy having tested Hardy in Vbox for some time. Vbox didnt work: tried to recompile it, reconfigure etc. Uninstalled 1.5 and downloaded the .deb for Hardy. Installed it and imported the WinXP guest.
All ok except for two problems which everyone is reporting on this and the ubuntu forums, with no solution thus far:
1: USB > the devices are recognised by VirtualBox, and can be seen when running the Windows Guest by right-clicking on the USB icon bottom right, but the USB devices themselves are greyed out and not selectable.
I've tried a range of usb devices, including those never used before. The result is consistent - Unavailable.
I've tried running Vbox as Root, and logging in to windows with the built in admin account - no dice.
I suspect it is partly a bug in VirtualBox and partly that Hardy (the host) is quite different from Gutsy in terms of security on devices.
2: Guest display size. The host has a lovely 1680x1050 resolution. Previously, VBox guest WinXP in "Full Screen" mode correctly did just that. Now, as widely reported by others, the max resolution available to the Guest is 1152x864 - and in full screen mode, there is a large amount of black border space around the guest.
It seems that there is a problem interpreting the Host available resolutions. Again, this might be a problem generated by the host OS changes - but it is not possible to test this against the previous version of vbox. Maybe when "updating" the product to include the Sun branding, something was inadvertantly buggered up?
So - any bright ideas?
NOTE: I've already run through the manual forwards, backwards, sideways. I've edited and unedited fstab, modprobe.d, and all of the other suggested config files that people have suggested.
My testing methodology is "make one change, then test the result before making another change"...
I hope by posting this that those who understand "under the covers" can indentify that there is definitely a bug to fix.