Friday, October 23, 2009

Spam in social software

From the Greg Linden's blog:

An arms race in spamming social software
by Greg Linden (16 Oct 2009)

The post presents an ultra-short, yet interesting, summary of spam strategies observed in social software. In particular, Greg Linden mentions several observed badmouthing strategies -- the spammer tries to taint competitors reputation, as opposed to attempt to use spam to obtain benefits directly. Interestingly, Chen and Friedman [1] conjecture that badmouthing attacks are harder to tackle than attacks that attempt to artificially boost one's reputation.

The post ends suggesting that incentive-based approaches to deter spam are (possibly good) alternatives compared to techniques that detect spam by focusing on content analysis (e.g., detection of commercial intent on blog comments). I would say they are complementary, though.

[1] Chen & Friedman. "Sybilproof Reputation Mechanisms".