Certainly host level security is important, whether data or voice. I am not saying that this should not be addressed, it's just that it is not a reason for not deploying VoIP. The fact that most of the exploits mentioned indicate that you already have a problem at the network boundary. If they are able to muck with the TFTP server, they can also screw up your DNS, wreck your router configs, divert all your VLANs etc, etc. Many of these systems are theoretically explotable, as they are protected by weak or non-existent security. Almost all WAN traffic (Frame Relay, ATM) is unencrypted. We *rely* on good border protection.
I mean no one "authenticates" analog (or digital PBX) phone systems today. The assumption is that there is "reasonable" building security and that the telcos infrastructure is hidden and secure. (However 100 metres from my house is the above-ground telco pillar which splits all the PSTN phone lines on my block. A couple of times every year kids remove the cover "for fun". This would be easily exploited). In a business environment, any "cleaner" or even a "visitor" could easily add phone taps that would not be easily detectable. (How about the infamous way you can configure a mobile phone to silently answer a call. Just leave it in a conference room, this is a bug that can be bought and used anywhere)
I would be interested to know whether there are any cases of people *actually* hacking/exploiting VoIP systems in real world environments. I also get involved in security audits and designs. While many exploits are technically possible, there is always a trade-off with cost and complexity when implementing security on systems. ---------------------------------------
I'm just trying to help, and am not a spokesman for HP