I've spent some time this morning clearing various junk messages out of various of the smaller social networks that I signed on for when they first launched. The only messages I had on them were those junk ones. All the messages had in common that they were from genuine-looking people that I didn't know.
I say smaller networks as they don't seem to have such good control over junk - they either have good processes in place, social tools to spot spammers or vast numbers of moderators. Twitter clearly makes good use of automation to spot inappropriate activity and had beaten down the wave of junk activity.
LinkedIn is an interesting conundrum as it effectively offers a very expensive license to spam. You can spam who you like if you pay enough, and a few companies actually do this. Typically offshore development or testing shops, I regularly get very polite messages somewhat crafted for my profile. Recognising that they come from people who are just doing a job that I've done too - selling software and services - I tend to take the time to explain I'm not a good target and won't be buying.
But the fact remains that the power of the network effect attracts those who spam and we don't yet have the tools to stop it automatically.