On one hand, I like the UPNP approach because it should only open what's necessary when it's necessary. On the other hand, it's a license for any rogue bit of malware on my network to open anything it wants. (Thankfully, I seldom run Windows.) If I get UPNP working, I should probably add firewall rules that allow only the XBOX to talk to miniupnpd.
This is a common misconception that doesn't stand up to analysis.
The fact is, if you have malware on your network, on a typical firewall it's fully capable of opening up any outbound connections it wants. UPnP does allow it to open up inbound ports too, but only in a limited way. Is there anything that can be done with a upnp inbound connection that couldn't, technically, be done through an outbound connection? No. In fact it's probably far easier and less likely to be detected (and certainly more reliable) for malware to create vulnerabilities through initiating outbound connections and local network sniffing.
The reality is in a lot of cases UPnP is a lot more secure than alternatives like static inbound mappings as the ports are only opened when required. They are also (if the upnp IGD is capable) loggable and monitorable.
Sure, you don't want UPnP on a typical corporate network, but there's certainly a big place for it on home networks and even SME networks.
Cheers,
Keith