That's a neat thought, but looking at the current pools: http://bitcoinchain.com/pools#start=1383289260&end=1386143940
you would need to gather > 2300Gh/s to even play in the field.
This wiki page: https://en.bitcoin.it/wiki/Mining_hardware_comparison#CPUs.2FAPUs
lists most CPUs as being capable of 2-30 Mh/s. Let's say an average CPU can do 10Mh/s at 100% utilization. But, we've gotta respect the user and only mine when they're away from their computer. Worst case is 33% of time (8 hours of sleep) Best case might be more like 66% (it's a work computer that gets used 8 hours a day). So, we need minimum 2300Gh/s / 6.6Mh/s ~= 350K users. If we had those users, that would have yielded 3 blocks over November at current rates, or ~ $86.6K.
That's pretty good, but it's not sustainable. That same amount of processing power would have yielded about 7 blocks in October, and about 23 blocks in September (looking at pools of equivalent power). That's a pretty sharp downward exponential which we would have to counter with a corresponding exponential increase in users.
In short, I think this would have once been a great strategy, but with the current rise in ASIC miners, I don't think a botnet (even a large one) is all that promising at the moment.