Technorati's David Sifry has published some statistics about the blogosphere, giving a round-up of current numbers and trends. Among the interesting facts in his post: 70% of 'pings' sent to Technorati to announce new blog postings come from known spam sources, and between 8% and 30% of new blogs created are 'splogs' (spam blogs). That 30% peak figure represents 50,000 blogs per hour. Spamming the blogosphere is a high-volume business.

The 70% figure is reminiscent of statistics for spam in other spheres. Current figures suggest that email spam may account for 85-90% of all email. More than 90% of the 35 million domains registered in April were apparently registered by spammers who never paid for them and simply allowed them to lapse after five days.

There's practically no aspect of current communication that doesn't have its own particular form of spam. Usenet spam - the 'original' spam - is now so pervasive that Usenet has become almost unusable. Spammers have developed dedicated tools for posting spam to bulletin boards and forums, while search engines such as Google are subject to continuous spamdexing attacks. The blogosphere is now home to comment spam, referrer spam, spings and splogs. Instant messaging and SMS spam is growing (recent figures for spim - instant messaging spam - are hard to find, but a study of spim in 2004 predicted that spim would account for 5% of instant messaging traffic by the end of the year), as is 'spit' - Internet telephony (VOIP) spam. There's even phishing on VOIP.

The bandwidth taken by spam is also rising. Recent months have seen an increase in image-based email spam, where the spam payload is carried as an embedded image rather than as text. A message that could previously be sent as a 3K text file might take 30K to send as an image. Currently, image-based spam accounts for around 20% of spam, but that figure seems likely to rise.

Delivering all these different types of spam requires significant amounts of infrastructure. The spammer's workhorse is the humble zombie PC, infected by a virus that turns it into a dedicated and configurable spam delivery engine. CipherTrust's zombie stats put the number of zombies detected daily at around 200,000. The messages that the zombies send out direct users to hundreds of thousands of websites - more than 70% of them hosted in China.

The irony is that the Internet begins to look, in essence, like a vast delivery engine for spam. For each new technology or communications channel, there's a new kind of spam adapted for delivery over that channel. In each case, spam often quickly expands to fill a majority of the resources available, to the point where just keeping the channel clear of spam so that the real traffic can be delivered accounts for a significant part of the cost of maintaining the channel.

The Internet was originally designed to resist external attacks, so that disabling a node or link in the network could not take down the whole network. The biggest challenge of the next few years may turn out to be hardening the network against the enemy within - designing an architecture that doesn't lend itself to colonization by abusive traffic.